diff --git a/frontend/src/App.tsx b/frontend/src/App.tsx
index 93a0a08..ba97090 100644
--- a/frontend/src/App.tsx
+++ b/frontend/src/App.tsx
@@ -4,6 +4,7 @@ import { DashboardOverview } from './components/DashboardOverview'
 import { Dashboard } from './components/Dashboard'
 import { DocumentDetail } from './components/DocumentDetail'
 import { Training } from './components/Training'
+import { DatasetDetail } from './components/DatasetDetail'
 import { Models } from './components/Models'
 import { Login } from './components/Login'
 import { InferenceDemo } from './components/InferenceDemo'
@@ -55,7 +56,14 @@ const App: React.FC = () => {
       case 'demo':
         return <InferenceDemo />
       case 'training':
-        return <Training />
+        return <Training onNavigate={handleNavigate} />
+      case 'dataset-detail':
+        return (
+          <DatasetDetail
+            datasetId={selectedDocId || ''}
+            onBack={() => setCurrentView('training')}
+          />
+        )
       case 'models':
         return <Models />
       default:
diff --git a/frontend/src/api/endpoints/augmentation.test.ts b/frontend/src/api/endpoints/augmentation.test.ts
new file mode 100644
index 0000000..9a0b7c2
--- /dev/null
+++ b/frontend/src/api/endpoints/augmentation.test.ts
@@ -0,0 +1,118 @@
+/**
+ * Tests for augmentation API endpoints.
+ *
+ * TDD Phase 1: RED - Write tests first, then implement to pass.
+ */
+
+import { describe, it, expect, vi, beforeEach } from 'vitest'
+import { augmentationApi } from './augmentation'
+import apiClient from '../client'
+
+// Mock the API client
+vi.mock('../client', () => ({
+  default: {
+    get: vi.fn(),
+    post: vi.fn(),
+  },
+}))
+
+describe('augmentationApi', () => {
+  beforeEach(() => {
+    vi.clearAllMocks()
+  })
+
+  describe('getTypes', () => {
+    it('should fetch augmentation types', async () => {
+      const mockResponse = {
+        data: {
+          augmentation_types: [
+            {
+              name: 'gaussian_noise',
+              description: 'Adds Gaussian noise',
+              affects_geometry: false,
+              stage: 'noise',
+              default_params: { mean: 0, std: 15 },
+            },
+          ],
+        },
+      }
+      vi.mocked(apiClient.get).mockResolvedValueOnce(mockResponse)
+
+      const result = await augmentationApi.getTypes()
+
+      expect(apiClient.get).toHaveBeenCalledWith('/api/v1/admin/augmentation/types')
+      expect(result.augmentation_types).toHaveLength(1)
+      expect(result.augmentation_types[0].name).toBe('gaussian_noise')
+    })
+  })
+
+  describe('getPresets', () => {
+    it('should fetch augmentation presets', async () => {
+      const mockResponse = {
+        data: {
+          presets: [
+            { name: 'conservative', description: 'Safe augmentations' },
+            { name: 'moderate', description: 'Balanced augmentations' },
+          ],
+        },
+      }
+      vi.mocked(apiClient.get).mockResolvedValueOnce(mockResponse)
+
+      const result = await augmentationApi.getPresets()
+
+      expect(apiClient.get).toHaveBeenCalledWith('/api/v1/admin/augmentation/presets')
+      expect(result.presets).toHaveLength(2)
+    })
+  })
+
+  describe('preview', () => {
+    it('should preview single augmentation', async () => {
+      const mockResponse = {
+        data: {
+          preview_url: 'data:image/png;base64,xxx',
+          original_url: 'data:image/png;base64,yyy',
+          applied_params: { std: 15 },
+        },
+      }
+      vi.mocked(apiClient.post).mockResolvedValueOnce(mockResponse)
+
+      const result = await augmentationApi.preview('doc-123', {
+        augmentation_type: 'gaussian_noise',
+        params: { std: 15 },
+      })
+
+      expect(apiClient.post).toHaveBeenCalledWith(
+        '/api/v1/admin/augmentation/preview/doc-123',
+        {
+          augmentation_type: 'gaussian_noise',
+          params: { std: 15 },
+        },
+        { params: { page: 1 } }
+      )
+      expect(result.preview_url).toBe('data:image/png;base64,xxx')
+    })
+
+    it('should support custom page number', async () => {
+      const mockResponse = {
+        data: {
+          preview_url: 'data:image/png;base64,xxx',
+          original_url: 'data:image/png;base64,yyy',
+          applied_params: {},
+        },
+      }
+      vi.mocked(apiClient.post).mockResolvedValueOnce(mockResponse)
+
+      await augmentationApi.preview(
+        'doc-123',
+        { augmentation_type: 'gaussian_noise', params: {} },
+        2
+      )
+
+      expect(apiClient.post).toHaveBeenCalledWith(
+        '/api/v1/admin/augmentation/preview/doc-123',
+        expect.anything(),
+        { params: { page: 2 } }
+      )
+    })
+  })
+})
diff --git a/frontend/src/api/endpoints/augmentation.ts b/frontend/src/api/endpoints/augmentation.ts
new file mode 100644
index 0000000..52cac76
--- /dev/null
+++ b/frontend/src/api/endpoints/augmentation.ts
@@ -0,0 +1,144 @@
+/**
+ * Augmentation API endpoints.
+ *
+ * Provides functions for fetching augmentation types, presets, and previewing augmentations.
+ */
+
+import apiClient from '../client'
+
+// Types
+export interface AugmentationTypeInfo {
+  name: string
+  description: string
+  affects_geometry: boolean
+  stage: string
+  default_params: Record<string, unknown>
+}
+
+export interface AugmentationTypesResponse {
+  augmentation_types: AugmentationTypeInfo[]
+}
+
+export interface PresetInfo {
+  name: string
+  description: string
+  config?: Record<string, unknown>
+}
+
+export interface PresetsResponse {
+  presets: PresetInfo[]
+}
+
+export interface PreviewRequest {
+  augmentation_type: string
+  params: Record<string, unknown>
+}
+
+export interface PreviewResponse {
+  preview_url: string
+  original_url: string
+  applied_params: Record<string, unknown>
+}
+
+export interface AugmentationParams {
+  enabled: boolean
+  probability: number
+  params: Record<string, unknown>
+}
+
+export interface AugmentationConfig {
+  perspective_warp?: AugmentationParams
+  wrinkle?: AugmentationParams
+  edge_damage?: AugmentationParams
+  stain?: AugmentationParams
+  lighting_variation?: AugmentationParams
+  shadow?: AugmentationParams
+  gaussian_blur?: AugmentationParams
+  motion_blur?: AugmentationParams
+  gaussian_noise?: AugmentationParams
+  salt_pepper?: AugmentationParams
+  paper_texture?: AugmentationParams
+  scanner_artifacts?: AugmentationParams
+  preserve_bboxes?: boolean
+  seed?: number | null
+}
+
+export interface BatchRequest {
+  dataset_id: string
+  config: AugmentationConfig
+  output_name: string
+  multiplier: number
+}
+
+export interface BatchResponse {
+  task_id: string
+  status: string
+  message: string
+  estimated_images: number
+}
+
+// API functions
+export const augmentationApi = {
+  /**
+   * Fetch available augmentation types.
+   */
+  async getTypes(): Promise<AugmentationTypesResponse> {
+    const response = await apiClient.get<AugmentationTypesResponse>(
+      '/api/v1/admin/augmentation/types'
+    )
+    return response.data
+  },
+
+  /**
+   * Fetch augmentation presets.
+   */
+  async getPresets(): Promise<PresetsResponse> {
+    const response = await apiClient.get<PresetsResponse>(
+      '/api/v1/admin/augmentation/presets'
+    )
+    return response.data
+  },
+
+  /**
+   * Preview a single augmentation on a document page.
+   */
+  async preview(
+    documentId: string,
+    request: PreviewRequest,
+    page: number = 1
+  ): Promise<PreviewResponse> {
+    const response = await apiClient.post<PreviewResponse>(
+      `/api/v1/admin/augmentation/preview/${documentId}`,
+      request,
+      { params: { page } }
+    )
+    return response.data
+  },
+
+  /**
+   * Preview full augmentation config on a document page.
+   */
+  async previewConfig(
+    documentId: string,
+    config: AugmentationConfig,
+    page: number = 1
+  ): Promise<PreviewResponse> {
+    const response = await apiClient.post<PreviewResponse>(
+      `/api/v1/admin/augmentation/preview-config/${documentId}`,
+      config,
+      { params: { page } }
+    )
+    return response.data
+  },
+
+  /**
+   * Create an augmented dataset.
+   */
+  async createBatch(request: BatchRequest): Promise<BatchResponse> {
+    const response = await apiClient.post<BatchResponse>(
+      '/api/v1/admin/augmentation/batch',
+      request
+    )
+    return response.data
+  },
+}
diff --git a/frontend/src/api/endpoints/datasets.ts b/frontend/src/api/endpoints/datasets.ts
new file mode 100644
index 0000000..42d1d67
--- /dev/null
+++ b/frontend/src/api/endpoints/datasets.ts
@@ -0,0 +1,52 @@
+import apiClient from '../client'
+import type {
+  DatasetCreateRequest,
+  DatasetDetailResponse,
+  DatasetListResponse,
+  DatasetResponse,
+  DatasetTrainRequest,
+  TrainingTaskResponse,
+} from '../types'
+
+export const datasetsApi = {
+  list: async (params?: {
+    status?: string
+    limit?: number
+    offset?: number
+  }): Promise<DatasetListResponse> => {
+    const { data } = await apiClient.get('/api/v1/admin/training/datasets', {
+      params,
+    })
+    return data
+  },
+
+  create: async (req: DatasetCreateRequest): Promise<DatasetResponse> => {
+    const { data } = await apiClient.post('/api/v1/admin/training/datasets', req)
+    return data
+  },
+
+  getDetail: async (datasetId: string): Promise<DatasetDetailResponse> => {
+    const { data } = await apiClient.get(
+      `/api/v1/admin/training/datasets/${datasetId}`
+    )
+    return data
+  },
+
+  remove: async (datasetId: string): Promise<{ message: string }> => {
+    const { data } = await apiClient.delete(
+      `/api/v1/admin/training/datasets/${datasetId}`
+    )
+    return data
+  },
+
+  trainFromDataset: async (
+    datasetId: string,
+    req: DatasetTrainRequest
+  ): Promise<TrainingTaskResponse> => {
+    const { data } = await apiClient.post(
+      `/api/v1/admin/training/datasets/${datasetId}/train`,
+      req
+    )
+    return data
+  },
+}
diff --git a/frontend/src/api/endpoints/documents.ts b/frontend/src/api/endpoints/documents.ts
index 56e5627..75367ed 100644
--- a/frontend/src/api/endpoints/documents.ts
+++ b/frontend/src/api/endpoints/documents.ts
@@ -21,14 +21,20 @@ export const documentsApi = {
     return data
   },
 
-  upload: async (file: File): Promise<UploadDocumentResponse> => {
+  upload: async (file: File, groupKey?: string): Promise<UploadDocumentResponse> => {
     const formData = new FormData()
     formData.append('file', file)
 
+    const params: Record<string, string> = {}
+    if (groupKey) {
+      params.group_key = groupKey
+    }
+
     const { data } = await apiClient.post('/api/v1/admin/documents', formData, {
       headers: {
         'Content-Type': 'multipart/form-data',
       },
+      params,
     })
     return data
   },
@@ -77,4 +83,16 @@ export const documentsApi = {
     )
     return data
   },
+
+  updateGroupKey: async (
+    documentId: string,
+    groupKey: string | null
+  ): Promise<{ status: string; document_id: string; group_key: string | null; message: string }> => {
+    const { data } = await apiClient.patch(
+      `/api/v1/admin/documents/${documentId}/group-key`,
+      null,
+      { params: { group_key: groupKey } }
+    )
+    return data
+  },
 }
diff --git a/frontend/src/api/endpoints/index.ts b/frontend/src/api/endpoints/index.ts
index f24f2f5..554ac30 100644
--- a/frontend/src/api/endpoints/index.ts
+++ b/frontend/src/api/endpoints/index.ts
@@ -2,3 +2,6 @@ export { documentsApi } from './documents'
 export { annotationsApi } from './annotations'
 export { trainingApi } from './training'
 export { inferenceApi } from './inference'
+export { datasetsApi } from './datasets'
+export { augmentationApi } from './augmentation'
+export { modelsApi } from './models'
diff --git a/frontend/src/api/endpoints/models.ts b/frontend/src/api/endpoints/models.ts
new file mode 100644
index 0000000..46b71c0
--- /dev/null
+++ b/frontend/src/api/endpoints/models.ts
@@ -0,0 +1,55 @@
+import apiClient from '../client'
+import type {
+  ModelVersionListResponse,
+  ModelVersionDetailResponse,
+  ModelVersionResponse,
+  ActiveModelResponse,
+} from '../types'
+
+export const modelsApi = {
+  list: async (params?: {
+    status?: string
+    limit?: number
+    offset?: number
+  }): Promise<ModelVersionListResponse> => {
+    const { data } = await apiClient.get('/api/v1/admin/training/models', {
+      params,
+    })
+    return data
+  },
+
+  getDetail: async (versionId: string): Promise<ModelVersionDetailResponse> => {
+    const { data } = await apiClient.get(`/api/v1/admin/training/models/${versionId}`)
+    return data
+  },
+
+  getActive: async (): Promise<ActiveModelResponse> => {
+    const { data } = await apiClient.get('/api/v1/admin/training/models/active')
+    return data
+  },
+
+  activate: async (versionId: string): Promise<ModelVersionResponse> => {
+    const { data } = await apiClient.post(`/api/v1/admin/training/models/${versionId}/activate`)
+    return data
+  },
+
+  deactivate: async (versionId: string): Promise<ModelVersionResponse> => {
+    const { data } = await apiClient.post(`/api/v1/admin/training/models/${versionId}/deactivate`)
+    return data
+  },
+
+  archive: async (versionId: string): Promise<ModelVersionResponse> => {
+    const { data } = await apiClient.post(`/api/v1/admin/training/models/${versionId}/archive`)
+    return data
+  },
+
+  delete: async (versionId: string): Promise<{ message: string }> => {
+    const { data } = await apiClient.delete(`/api/v1/admin/training/models/${versionId}`)
+    return data
+  },
+
+  reload: async (): Promise<{ message: string; reloaded: boolean }> => {
+    const { data } = await apiClient.post('/api/v1/admin/training/models/reload')
+    return data
+  },
+}
diff --git a/frontend/src/api/types.ts b/frontend/src/api/types.ts
index c668a59..73908ca 100644
--- a/frontend/src/api/types.ts
+++ b/frontend/src/api/types.ts
@@ -8,6 +8,7 @@ export interface DocumentItem {
   auto_label_status: 'pending' | 'running' | 'completed' | 'failed' | null
   auto_label_error: string | null
   upload_source: string
+  group_key: string | null
   created_at: string
   updated_at: string
   annotation_count?: number
@@ -59,6 +60,7 @@ export interface DocumentDetailResponse {
   auto_label_error: string | null
   upload_source: string
   batch_id: string | null
+  group_key: string | null
   csv_field_values: Record<string, string> | null
   can_annotate: boolean
   annotation_lock_until: string | null
@@ -113,7 +115,11 @@ export interface ErrorResponse {
 export interface UploadDocumentResponse {
   document_id: string
   filename: string
+  file_size: number
+  page_count: number
   status: string
+  group_key: string | null
+  auto_label_started: boolean
   message: string
 }
 
@@ -171,3 +177,165 @@ export interface InferenceResult {
 export interface InferenceResponse {
   result: InferenceResult
 }
+
+// Dataset types
+
+export interface DatasetCreateRequest {
+  name: string
+  description?: string
+  document_ids: string[]
+  train_ratio?: number
+  val_ratio?: number
+  seed?: number
+}
+
+export interface DatasetResponse {
+  dataset_id: string
+  name: string
+  status: string
+  message: string
+}
+
+export interface DatasetDocumentItem {
+  document_id: string
+  split: string
+  page_count: number
+  annotation_count: number
+}
+
+export interface DatasetListItem {
+  dataset_id: string
+  name: string
+  description: string | null
+  status: string
+  training_status: string | null
+  active_training_task_id: string | null
+  total_documents: number
+  total_images: number
+  total_annotations: number
+  created_at: string
+}
+
+export interface DatasetListResponse {
+  total: number
+  limit: number
+  offset: number
+  datasets: DatasetListItem[]
+}
+
+export interface DatasetDetailResponse {
+  dataset_id: string
+  name: string
+  description: string | null
+  status: string
+  train_ratio: number
+  val_ratio: number
+  seed: number
+  total_documents: number
+  total_images: number
+  total_annotations: number
+  dataset_path: string | null
+  error_message: string | null
+  documents: DatasetDocumentItem[]
+  created_at: string
+  updated_at: string
+}
+
+export interface AugmentationParams {
+  enabled: boolean
+  probability: number
+  params: Record<string, unknown>
+}
+
+export interface AugmentationTrainingConfig {
+  gaussian_noise?: AugmentationParams
+  perspective_warp?: AugmentationParams
+  wrinkle?: AugmentationParams
+  edge_damage?: AugmentationParams
+  stain?: AugmentationParams
+  lighting_variation?: AugmentationParams
+  shadow?: AugmentationParams
+  gaussian_blur?: AugmentationParams
+  motion_blur?: AugmentationParams
+  salt_pepper?: AugmentationParams
+  paper_texture?: AugmentationParams
+  scanner_artifacts?: AugmentationParams
+  preserve_bboxes?: boolean
+  seed?: number | null
+}
+
+export interface DatasetTrainRequest {
+  name: string
+  config: {
+    model_name?: string
+    base_model_version_id?: string | null
+    epochs?: number
+    batch_size?: number
+    image_size?: number
+    learning_rate?: number
+    device?: string
+    augmentation?: AugmentationTrainingConfig
+    augmentation_multiplier?: number
+  }
+}
+
+export interface TrainingTaskResponse {
+  task_id: string
+  status: string
+  message: string
+}
+
+// Model Version types
+
+export interface ModelVersionItem {
+  version_id: string
+  version: string
+  name: string
+  status: string
+  is_active: boolean
+  metrics_mAP: number | null
+  document_count: number
+  trained_at: string | null
+  activated_at: string | null
+  created_at: string
+}
+
+export interface ModelVersionDetailResponse {
+  version_id: string
+  version: string
+  name: string
+  description: string | null
+  model_path: string
+  status: string
+  is_active: boolean
+  task_id: string | null
+  dataset_id: string | null
+  metrics_mAP: number | null
+  metrics_precision: number | null
+  metrics_recall: number | null
+  document_count: number
+  training_config: Record<string, unknown> | null
+  file_size: number | null
+  trained_at: string | null
+  activated_at: string | null
+  created_at: string
+  updated_at: string
+}
+
+export interface ModelVersionListResponse {
+  total: number
+  limit: number
+  offset: number
+  models: ModelVersionItem[]
+}
+
+export interface ModelVersionResponse {
+  version_id: string
+  status: string
+  message: string
+}
+
+export interface ActiveModelResponse {
+  has_active_model: boolean
+  model: ModelVersionItem | null
+}
diff --git a/frontend/src/components/AugmentationConfig.test.tsx b/frontend/src/components/AugmentationConfig.test.tsx
new file mode 100644
index 0000000..f79a1be
--- /dev/null
+++ b/frontend/src/components/AugmentationConfig.test.tsx
@@ -0,0 +1,251 @@
+/**
+ * Tests for AugmentationConfig component.
+ *
+ * TDD Phase 1: RED - Write tests first, then implement to pass.
+ */
+
+import { describe, it, expect, vi, beforeEach } from 'vitest'
+import { render, screen, fireEvent, waitFor } from '@testing-library/react'
+import userEvent from '@testing-library/user-event'
+import { QueryClient, QueryClientProvider } from '@tanstack/react-query'
+import { AugmentationConfig } from './AugmentationConfig'
+import { augmentationApi } from '../api/endpoints/augmentation'
+import type { ReactNode } from 'react'
+
+// Mock the API
+vi.mock('../api/endpoints/augmentation', () => ({
+  augmentationApi: {
+    getTypes: vi.fn(),
+    getPresets: vi.fn(),
+    preview: vi.fn(),
+    previewConfig: vi.fn(),
+    createBatch: vi.fn(),
+  },
+}))
+
+// Default mock data
+const mockTypes = {
+  augmentation_types: [
+    {
+      name: 'gaussian_noise',
+      description: 'Adds Gaussian noise to simulate sensor noise',
+      affects_geometry: false,
+      stage: 'noise',
+      default_params: { mean: 0, std: 15 },
+    },
+    {
+      name: 'perspective_warp',
+      description: 'Applies perspective transformation',
+      affects_geometry: true,
+      stage: 'geometric',
+      default_params: { max_warp: 0.02 },
+    },
+    {
+      name: 'gaussian_blur',
+      description: 'Applies Gaussian blur',
+      affects_geometry: false,
+      stage: 'blur',
+      default_params: { kernel_size: 5 },
+    },
+  ],
+}
+
+const mockPresets = {
+  presets: [
+    { name: 'conservative', description: 'Safe augmentations for high-quality documents' },
+    { name: 'moderate', description: 'Balanced augmentation settings' },
+    { name: 'aggressive', description: 'Strong augmentations for data diversity' },
+  ],
+}
+
+// Test wrapper with QueryClient
+const createWrapper = () => {
+  const queryClient = new QueryClient({
+    defaultOptions: {
+      queries: {
+        retry: false,
+      },
+    },
+  })
+  return ({ children }: { children: ReactNode }) => (
+    <QueryClientProvider client={queryClient}>{children}</QueryClientProvider>
+  )
+}
+
+describe('AugmentationConfig', () => {
+  beforeEach(() => {
+    vi.clearAllMocks()
+    vi.mocked(augmentationApi.getTypes).mockResolvedValue(mockTypes)
+    vi.mocked(augmentationApi.getPresets).mockResolvedValue(mockPresets)
+  })
+
+  describe('rendering', () => {
+    it('should render enable checkbox', async () => {
+      render(
+        <AugmentationConfig
+          enabled={false}
+          onEnabledChange={vi.fn()}
+          config={{}}
+          onConfigChange={vi.fn()}
+        />,
+        { wrapper: createWrapper() }
+      )
+
+      expect(screen.getByRole('checkbox', { name: /enable augmentation/i })).toBeInTheDocument()
+    })
+
+    it('should be collapsed when disabled', () => {
+      render(
+        <AugmentationConfig
+          enabled={false}
+          onEnabledChange={vi.fn()}
+          config={{}}
+          onConfigChange={vi.fn()}
+        />,
+        { wrapper: createWrapper() }
+      )
+
+      // Config options should not be visible
+      expect(screen.queryByText(/preset/i)).not.toBeInTheDocument()
+    })
+
+    it('should expand when enabled', async () => {
+      render(
+        <AugmentationConfig
+          enabled={true}
+          onEnabledChange={vi.fn()}
+          config={{}}
+          onConfigChange={vi.fn()}
+        />,
+        { wrapper: createWrapper() }
+      )
+
+      await waitFor(() => {
+        expect(screen.getByText(/preset/i)).toBeInTheDocument()
+      })
+    })
+  })
+
+  describe('preset selection', () => {
+    it('should display available presets', async () => {
+      render(
+        <AugmentationConfig
+          enabled={true}
+          onEnabledChange={vi.fn()}
+          config={{}}
+          onConfigChange={vi.fn()}
+        />,
+        { wrapper: createWrapper() }
+      )
+
+      await waitFor(() => {
+        expect(screen.getByText('conservative')).toBeInTheDocument()
+        expect(screen.getByText('moderate')).toBeInTheDocument()
+        expect(screen.getByText('aggressive')).toBeInTheDocument()
+      })
+    })
+
+    it('should call onConfigChange when preset is selected', async () => {
+      const user = userEvent.setup()
+      const onConfigChange = vi.fn()
+
+      render(
+        <AugmentationConfig
+          enabled={true}
+          onEnabledChange={vi.fn()}
+          config={{}}
+          onConfigChange={onConfigChange}
+        />,
+        { wrapper: createWrapper() }
+      )
+
+      await waitFor(() => {
+        expect(screen.getByText('moderate')).toBeInTheDocument()
+      })
+
+      await user.click(screen.getByText('moderate'))
+
+      expect(onConfigChange).toHaveBeenCalled()
+    })
+  })
+
+  describe('enable toggle', () => {
+    it('should call onEnabledChange when checkbox is toggled', async () => {
+      const user = userEvent.setup()
+      const onEnabledChange = vi.fn()
+
+      render(
+        <AugmentationConfig
+          enabled={false}
+          onEnabledChange={onEnabledChange}
+          config={{}}
+          onConfigChange={vi.fn()}
+        />,
+        { wrapper: createWrapper() }
+      )
+
+      await user.click(screen.getByRole('checkbox', { name: /enable augmentation/i }))
+
+      expect(onEnabledChange).toHaveBeenCalledWith(true)
+    })
+  })
+
+  describe('augmentation types', () => {
+    it('should display augmentation types when in custom mode', async () => {
+      render(
+        <AugmentationConfig
+          enabled={true}
+          onEnabledChange={vi.fn()}
+          config={{}}
+          onConfigChange={vi.fn()}
+          showCustomOptions={true}
+        />,
+        { wrapper: createWrapper() }
+      )
+
+      await waitFor(() => {
+        expect(screen.getByText(/gaussian_noise/i)).toBeInTheDocument()
+        expect(screen.getByText(/perspective_warp/i)).toBeInTheDocument()
+      })
+    })
+
+    it('should indicate which augmentations affect geometry', async () => {
+      render(
+        <AugmentationConfig
+          enabled={true}
+          onEnabledChange={vi.fn()}
+          config={{}}
+          onConfigChange={vi.fn()}
+          showCustomOptions={true}
+        />,
+        { wrapper: createWrapper() }
+      )
+
+      await waitFor(() => {
+        // perspective_warp affects geometry
+        const perspectiveItem = screen.getByText(/perspective_warp/i).closest('div')
+        expect(perspectiveItem).toHaveTextContent(/affects bbox/i)
+      })
+    })
+  })
+
+  describe('loading state', () => {
+    it('should show loading indicator while fetching types', () => {
+      vi.mocked(augmentationApi.getTypes).mockImplementation(
+        () => new Promise(() => {})
+      )
+
+      render(
+        <AugmentationConfig
+          enabled={true}
+          onEnabledChange={vi.fn()}
+          config={{}}
+          onConfigChange={vi.fn()}
+        />,
+        { wrapper: createWrapper() }
+      )
+
+      expect(screen.getByTestId('augmentation-loading')).toBeInTheDocument()
+    })
+  })
+})
diff --git a/frontend/src/components/AugmentationConfig.tsx b/frontend/src/components/AugmentationConfig.tsx
new file mode 100644
index 0000000..07ffdd6
--- /dev/null
+++ b/frontend/src/components/AugmentationConfig.tsx
@@ -0,0 +1,136 @@
+/**
+ * AugmentationConfig component for configuring image augmentation during training.
+ *
+ * Provides preset selection and optional custom augmentation type configuration.
+ */
+
+import React from 'react'
+import { Loader2, AlertTriangle } from 'lucide-react'
+import { useAugmentation } from '../hooks/useAugmentation'
+import type { AugmentationConfig as AugmentationConfigType } from '../api/endpoints/augmentation'
+
+interface AugmentationConfigProps {
+  enabled: boolean
+  onEnabledChange: (enabled: boolean) => void
+  config: Partial<AugmentationConfigType>
+  onConfigChange: (config: Partial<AugmentationConfigType>) => void
+  showCustomOptions?: boolean
+}
+
+export const AugmentationConfig: React.FC<AugmentationConfigProps> = ({
+  enabled,
+  onEnabledChange,
+  config,
+  onConfigChange,
+  showCustomOptions = false,
+}) => {
+  const { augmentationTypes, presets, isLoadingTypes, isLoadingPresets } = useAugmentation()
+
+  const isLoading = isLoadingTypes || isLoadingPresets
+
+  const handlePresetSelect = (presetName: string) => {
+    const preset = presets.find((p) => p.name === presetName)
+    if (preset && preset.config) {
+      onConfigChange(preset.config as Partial<AugmentationConfigType>)
+    } else {
+      // Apply a basic config based on preset name
+      const presetConfigs: Record<string, Partial<AugmentationConfigType>> = {
+        conservative: {
+          gaussian_noise: { enabled: true, probability: 0.3, params: { std: 10 } },
+          gaussian_blur: { enabled: true, probability: 0.2, params: { kernel_size: 3 } },
+        },
+        moderate: {
+          gaussian_noise: { enabled: true, probability: 0.5, params: { std: 15 } },
+          gaussian_blur: { enabled: true, probability: 0.3, params: { kernel_size: 5 } },
+          lighting_variation: { enabled: true, probability: 0.3, params: {} },
+          perspective_warp: { enabled: true, probability: 0.2, params: { max_warp: 0.02 } },
+        },
+        aggressive: {
+          gaussian_noise: { enabled: true, probability: 0.7, params: { std: 20 } },
+          gaussian_blur: { enabled: true, probability: 0.5, params: { kernel_size: 7 } },
+          motion_blur: { enabled: true, probability: 0.3, params: {} },
+          lighting_variation: { enabled: true, probability: 0.5, params: {} },
+          shadow: { enabled: true, probability: 0.3, params: {} },
+          perspective_warp: { enabled: true, probability: 0.3, params: { max_warp: 0.03 } },
+          wrinkle: { enabled: true, probability: 0.2, params: {} },
+          stain: { enabled: true, probability: 0.2, params: {} },
+        },
+      }
+      onConfigChange(presetConfigs[presetName] || {})
+    }
+  }
+
+  return (
+    <div className="border border-warm-divider rounded-lg p-4 bg-warm-bg-secondary">
+      {/* Enable checkbox */}
+      <label className="flex items-center gap-2 cursor-pointer">
+        <input
+          type="checkbox"
+          checked={enabled}
+          onChange={(e) => onEnabledChange(e.target.checked)}
+          className="w-4 h-4 rounded border-warm-divider text-warm-state-info focus:ring-warm-state-info"
+          aria-label="Enable augmentation"
+        />
+        <span className="text-sm font-medium text-warm-text-secondary">Enable Augmentation</span>
+        <span className="text-xs text-warm-text-muted">(Simulate real-world document conditions)</span>
+      </label>
+
+      {/* Expanded content when enabled */}
+      {enabled && (
+        <div className="mt-4 space-y-4">
+          {isLoading ? (
+            <div className="flex items-center justify-center py-4" data-testid="augmentation-loading">
+              <Loader2 className="w-5 h-5 animate-spin text-warm-state-info" />
+              <span className="ml-2 text-sm text-warm-text-muted">Loading augmentation options...</span>
+            </div>
+          ) : (
+            <>
+              {/* Preset selection */}
+              <div>
+                <label className="block text-sm font-medium text-warm-text-secondary mb-2">Preset</label>
+                <div className="flex flex-wrap gap-2">
+                  {presets.map((preset) => (
+                    <button
+                      key={preset.name}
+                      onClick={() => handlePresetSelect(preset.name)}
+                      className="px-3 py-1.5 text-sm rounded-md border border-warm-divider hover:bg-warm-bg-tertiary transition-colors"
+                      title={preset.description}
+                    >
+                      {preset.name}
+                    </button>
+                  ))}
+                </div>
+              </div>
+
+              {/* Custom options (if enabled) */}
+              {showCustomOptions && (
+                <div className="border-t border-warm-divider pt-4">
+                  <h4 className="text-sm font-medium text-warm-text-secondary mb-3">Augmentation Types</h4>
+                  <div className="grid gap-2">
+                    {augmentationTypes.map((type) => (
+                      <div
+                        key={type.name}
+                        className="flex items-center justify-between p-2 bg-warm-bg-primary rounded border border-warm-divider"
+                      >
+                        <div className="flex items-center gap-2">
+                          <span className="text-sm text-warm-text-primary">{type.name}</span>
+                          {type.affects_geometry && (
+                            <span className="flex items-center gap-1 text-xs text-warm-state-warning">
+                              <AlertTriangle size={12} />
+                              affects bbox
+                            </span>
+                          )}
+                        </div>
+                        <span className="text-xs text-warm-text-muted">{type.stage}</span>
+                      </div>
+                    ))}
+                  </div>
+                </div>
+              )}
+            </>
+          )}
+        </div>
+      )}
+    </div>
+  )
+}
diff --git a/frontend/src/components/Dashboard.tsx b/frontend/src/components/Dashboard.tsx
index 02517d7..5572734 100644
--- a/frontend/src/components/Dashboard.tsx
+++ b/frontend/src/components/Dashboard.tsx
@@ -144,6 +144,9 @@ export const Dashboard: React.FC<DashboardProps> = ({ onNavigate }) => {
               <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase tracking-wider">
                 Annotations
               </th>
+              <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase tracking-wider">
+                Group
+              </th>
               <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase tracking-wider w-64">
                 Auto-label
               </th>
@@ -153,13 +156,13 @@ export const Dashboard: React.FC<DashboardProps> = ({ onNavigate }) => {
           <tbody>
             {isLoading ? (
               <tr>
-                <td colSpan={7} className="py-8 text-center text-warm-text-muted">
+                <td colSpan={8} className="py-8 text-center text-warm-text-muted">
                   Loading documents...
                 </td>
               </tr>
             ) : documents.length === 0 ? (
               <tr>
-                <td colSpan={7} className="py-8 text-center text-warm-text-muted">
+                <td colSpan={8} className="py-8 text-center text-warm-text-muted">
                   No documents found. Upload your first document to get started.
                 </td>
               </tr>
@@ -213,6 +216,9 @@ export const Dashboard: React.FC<DashboardProps> = ({ onNavigate }) => {
                     <td className="py-4 px-4 text-sm text-warm-text-secondary">
                       {doc.annotation_count || 0} annotations
                     </td>
+                    <td className="py-4 px-4 text-sm text-warm-text-muted">
+                      {doc.group_key || '-'}
+                    </td>
                     <td className="py-4 px-4">
                       {doc.auto_label_status === 'running' && progress && (
                         <div className="w-full">
diff --git a/frontend/src/components/DatasetDetail.tsx b/frontend/src/components/DatasetDetail.tsx
new file mode 100644
index 0000000..b70b704
--- /dev/null
+++ b/frontend/src/components/DatasetDetail.tsx
@@ -0,0 +1,122 @@
+import React from 'react'
+import { ArrowLeft, Loader2, Play, AlertCircle, Check } from 'lucide-react'
+import { Button } from './Button'
+import { useDatasetDetail } from '../hooks/useDatasets'
+
+interface DatasetDetailProps {
+  datasetId: string
+  onBack: () => void
+}
+
+const SPLIT_STYLES: Record<string, string> = {
+  train: 'bg-warm-state-info/10 text-warm-state-info',
+  val: 'bg-warm-state-warning/10 text-warm-state-warning',
+  test: 'bg-warm-state-success/10 text-warm-state-success',
+}
+
+export const DatasetDetail: React.FC<DatasetDetailProps> = ({ datasetId, onBack }) => {
+  const { dataset, isLoading, error } = useDatasetDetail(datasetId)
+
+  if (isLoading) {
+    return (
+      <div className="flex items-center justify-center py-20 text-warm-text-muted">
+        <Loader2 size={24} className="animate-spin mr-2" />Loading dataset...
+      </div>
+    )
+  }
+
+  if (error || !dataset) {
+    return (
+      <div className="p-8 max-w-7xl mx-auto">
+        <button onClick={onBack} className="flex items-center gap-1 text-sm text-warm-text-muted hover:text-warm-text-secondary mb-4">
+          <ArrowLeft size={16} />Back
+        </button>
+        <p className="text-warm-state-error">Failed to load dataset.</p>
+      </div>
+    )
+  }
+
+  const statusIcon = dataset.status === 'ready'
+    ? <Check size={14} className="text-warm-state-success" />
+    : dataset.status === 'failed'
+    ? <AlertCircle size={14} className="text-warm-state-error" />
+    : <Loader2 size={14} className="animate-spin text-warm-state-info" />
+
+  return (
+    <div className="p-8 max-w-7xl mx-auto">
+      {/* Header */}
+      <button onClick={onBack} className="flex items-center gap-1 text-sm text-warm-text-muted hover:text-warm-text-secondary mb-4">
+        <ArrowLeft size={16} />Back to Datasets
+      </button>
+
+      <div className="flex items-center justify-between mb-6">
+        <div>
+          <h2 className="text-2xl font-bold text-warm-text-primary flex items-center gap-2">
+            {dataset.name} {statusIcon}
+          </h2>
+          {dataset.description && (
+            <p className="text-sm text-warm-text-muted mt-1">{dataset.description}</p>
+          )}
+        </div>
+        {dataset.status === 'ready' && (
+          <Button><Play size={14} className="mr-1" />Start Training</Button>
+        )}
+      </div>
+
+      {dataset.error_message && (
+        <div className="bg-warm-state-error/10 border border-warm-state-error/20 rounded-lg p-4 mb-6 text-sm text-warm-state-error">
+          {dataset.error_message}
+        </div>
+      )}
+
+      {/* Stats */}
+      <div className="grid grid-cols-4 gap-4 mb-8">
+        {[
+          ['Documents', dataset.total_documents],
+          ['Images', dataset.total_images],
+          ['Annotations', dataset.total_annotations],
+          ['Split', `${(dataset.train_ratio * 100).toFixed(0)}/${(dataset.val_ratio * 100).toFixed(0)}/${((1 - dataset.train_ratio - dataset.val_ratio) * 100).toFixed(0)}`],
+        ].map(([label, value]) => (
+          <div key={String(label)} className="bg-warm-card border border-warm-border rounded-lg p-4">
+            <p className="text-xs text-warm-text-muted uppercase font-semibold mb-1">{label}</p>
+            <p className="text-2xl font-bold text-warm-text-primary font-mono">{value}</p>
+          </div>
+        ))}
+      </div>
+
+      {/* Document list */}
+      <h3 className="text-lg font-semibold text-warm-text-primary mb-4">Documents</h3>
+      <div className="bg-warm-card border border-warm-border rounded-lg overflow-hidden shadow-sm">
+        <table className="w-full text-left">
+          <thead className="bg-white border-b border-warm-border">
+            <tr>
+              <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase">Document ID</th>
+              <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase">Split</th>
+              <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase">Pages</th>
+              <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase">Annotations</th>
+            </tr>
+          </thead>
+          <tbody>
+            {dataset.documents.map(doc => (
+              <tr key={doc.document_id} className="border-b border-warm-border hover:bg-warm-hover transition-colors">
+                <td className="py-3 px-4 text-sm font-mono text-warm-text-secondary">{doc.document_id.slice(0, 8)}...</td>
+                <td className="py-3 px-4">
+                  <span className={`inline-flex px-2.5 py-1 rounded-full text-xs font-medium ${SPLIT_STYLES[doc.split] ?? 'bg-warm-border text-warm-text-muted'}`}>
+                    {doc.split}
+                  </span>
+                </td>
+                <td className="py-3 px-4 text-sm text-warm-text-muted font-mono">{doc.page_count}</td>
+                <td className="py-3 px-4 text-sm text-warm-text-muted font-mono">{doc.annotation_count}</td>
+              </tr>
+            ))}
+          </tbody>
+        </table>
+      </div>
+
+      <p className="text-xs text-warm-text-muted mt-4">
+        Created: {new Date(dataset.created_at).toLocaleString()} | Updated: {new Date(dataset.updated_at).toLocaleString()}
+        {dataset.dataset_path && <> | Path: <code className="text-xs">{dataset.dataset_path}</code></>}
+      </p>
+    </div>
+  )
+}
diff --git a/frontend/src/components/DocumentDetail.tsx b/frontend/src/components/DocumentDetail.tsx
index cf4da73..258a4ba 100644
--- a/frontend/src/components/DocumentDetail.tsx
+++ b/frontend/src/components/DocumentDetail.tsx
@@ -1,8 +1,9 @@
 import React, { useState, useRef, useEffect } from 'react'
-import { ChevronLeft, ZoomIn, ZoomOut, Plus, Edit2, Trash2, Tag, CheckCircle } from 'lucide-react'
+import { ChevronLeft, ZoomIn, ZoomOut, Plus, Edit2, Trash2, Tag, CheckCircle, Check, X } from 'lucide-react'
 import { Button } from './Button'
 import { useDocumentDetail } from '../hooks/useDocumentDetail'
 import { useAnnotations } from '../hooks/useAnnotations'
+import { useDocuments } from '../hooks/useDocuments'
 import { documentsApi } from '../api/endpoints/documents'
 import type { AnnotationItem } from '../api/types'
 
@@ -26,7 +27,7 @@ const FIELD_CLASSES: Record<number, string> = {
 }
 
 export const DocumentDetail: React.FC<DocumentDetailProps> = ({ docId, onBack }) => {
-  const { document, annotations, isLoading } = useDocumentDetail(docId)
+  const { document, annotations, isLoading, refetch } = useDocumentDetail(docId)
   const {
     createAnnotation,
     updateAnnotation,
@@ -34,10 +35,13 @@ export const DocumentDetail: React.FC<DocumentDetailProps> = ({ docId, onBack })
     isCreating,
     isDeleting,
   } = useAnnotations(docId)
+  const { updateGroupKey, isUpdatingGroupKey } = useDocuments({})
 
   const [selectedId, setSelectedId] = useState<string | null>(null)
   const [zoom, setZoom] = useState(100)
   const [isDrawing, setIsDrawing] = useState(false)
+  const [isEditingGroupKey, setIsEditingGroupKey] = useState(false)
+  const [editGroupKeyValue, setEditGroupKeyValue] = useState('')
   const [drawStart, setDrawStart] = useState<{ x: number; y: number } | null>(null)
   const [drawEnd, setDrawEnd] = useState<{ x: number; y: number } | null>(null)
   const [selectedClassId, setSelectedClassId] = useState<number>(0)
@@ -426,6 +430,65 @@ export const DocumentDetail: React.FC<DocumentDetailProps> = ({ docId, onBack })
                   {new Date(document.created_at).toLocaleDateString()}
                 </span>
               </div>
+              <div className="flex justify-between items-center text-xs">
+                <span className="text-warm-text-muted">Group</span>
+                {isEditingGroupKey ? (
+                  <div className="flex items-center gap-1">
+                    <input
+                      type="text"
+                      value={editGroupKeyValue}
+                      onChange={(e) => setEditGroupKeyValue(e.target.value)}
+                      className="w-24 px-1.5 py-0.5 text-xs border border-warm-border rounded focus:outline-none focus:ring-1 focus:ring-warm-state-info"
+                      placeholder="group key"
+                      autoFocus
+                    />
+                    <button
+                      onClick={() => {
+                        updateGroupKey(
+                          { documentId: docId, groupKey: editGroupKeyValue.trim() || null },
+                          {
+                            onSuccess: () => {
+                              setIsEditingGroupKey(false)
+                              refetch()
+                            },
+                            onError: () => {
+                              alert('Failed to update group key. Please try again.')
+                            },
+                          }
+                        )
+                      }}
+                      disabled={isUpdatingGroupKey}
+                      className="p-0.5 text-warm-state-success hover:bg-warm-hover rounded"
+                    >
+                      <Check size={14} />
+                    </button>
+                    <button
+                      onClick={() => {
+                        setIsEditingGroupKey(false)
+                        setEditGroupKeyValue(document.group_key || '')
+                      }}
+                      className="p-0.5 text-warm-state-error hover:bg-warm-hover rounded"
+                    >
+                      <X size={14} />
+                    </button>
+                  </div>
+                ) : (
+                  <div className="flex items-center gap-1">
+                    <span className="text-warm-text-secondary font-medium">
+                      {document.group_key || '-'}
+                    </span>
+                    <button
+                      onClick={() => {
+                        setEditGroupKeyValue(document.group_key || '')
+                        setIsEditingGroupKey(true)
+                      }}
+                      className="p-0.5 text-warm-text-muted hover:text-warm-text-secondary hover:bg-warm-hover rounded"
+                    >
+                      <Edit2 size={12} />
+                    </button>
+                  </div>
+                )}
+              </div>
             </div>
           </div>
         </div>
diff --git a/frontend/src/components/Models.tsx b/frontend/src/components/Models.tsx
index c35052f..bfe2222 100644
--- a/frontend/src/components/Models.tsx
+++ b/frontend/src/components/Models.tsx
@@ -1,73 +1,108 @@
-import React from 'react';
+import React, { useState } from 'react';
 import { BarChart, Bar, XAxis, YAxis, CartesianGrid, Tooltip, ResponsiveContainer } from 'recharts';
+import { Loader2, Power, CheckCircle } from 'lucide-react';
 import { Button } from './Button';
+import { useModels, useModelDetail } from '../hooks';
+import type { ModelVersionItem } from '../api/types';
 
-const CHART_DATA = [
-  { name: 'Model A', value: 75 },
-  { name: 'Model B', value: 82 },
-  { name: 'Model C', value: 95 },
-  { name: 'Model D', value: 68 },
-];
-
-const METRICS_DATA = [
-  { name: 'Precision', value: 88 },
-  { name: 'Recall', value: 76 },
-  { name: 'F1 Score', value: 91 },
-  { name: 'Accuracy', value: 82 },
-];
-
-const JOBS = [
-  { id: 1, name: 'Training Job Job 1', date: '12/29/2024 10:33 PM', status: 'Running', progress: 65 },
-  { id: 2, name: 'Training Job 2', date: '12/29/2024 10:33 PM', status: 'Completed', success: 37, metrics: 89 },
-  { id: 3, name: 'Model Training Compentr 1', date: '12/29/2024 10:19 PM', status: 'Completed', success: 87, metrics: 92 },
-];
+const formatDate = (dateString: string | null): string => {
+  if (!dateString) return 'N/A';
+  return new Date(dateString).toLocaleString();
+};
 
 export const Models: React.FC = () => {
+  const [selectedModel, setSelectedModel] = useState<ModelVersionItem | null>(null);
+  const { models, isLoading, activateModel, isActivating } = useModels();
+  const { model: modelDetail } = useModelDetail(selectedModel?.version_id ?? null);
+
+  // Build chart data from selected model's metrics
+  const metricsData = modelDetail ? [
+    { name: 'Precision', value: (modelDetail.metrics_precision ?? 0) * 100 },
+    { name: 'Recall', value: (modelDetail.metrics_recall ?? 0) * 100 },
+    { name: 'mAP', value: (modelDetail.metrics_mAP ?? 0) * 100 },
+  ] : [
+    { name: 'Precision', value: 0 },
+    { name: 'Recall', value: 0 },
+    { name: 'mAP', value: 0 },
+  ];
+
+  // Build comparison chart from all models (with placeholder if empty)
+  const chartData = models.length > 0
+    ? models.slice(0, 4).map(m => ({
+        name: m.version,
+        value: (m.metrics_mAP ?? 0) * 100,
+      }))
+    : [
+        { name: 'Model A', value: 0 },
+        { name: 'Model B', value: 0 },
+        { name: 'Model C', value: 0 },
+        { name: 'Model D', value: 0 },
+      ];
+
   return (
     <div className="p-8 max-w-7xl mx-auto flex gap-8">
       {/* Left: Job History */}
       <div className="flex-1">
         <h2 className="text-2xl font-bold text-warm-text-primary mb-6">Models & History</h2>
-        <h3 className="text-lg font-semibold text-warm-text-primary mb-4">Recent Training Jobs</h3>
+        <h3 className="text-lg font-semibold text-warm-text-primary mb-4">Model Versions</h3>
 
-        <div className="space-y-4">
-          {JOBS.map(job => (
-            <div key={job.id} className="bg-warm-card border border-warm-border rounded-lg p-5 shadow-sm hover:border-warm-divider transition-colors">
-              <div className="flex justify-between items-start mb-2">
-                 <div>
-                   <h4 className="font-semibold text-warm-text-primary text-lg mb-1">{job.name}</h4>
-                   <p className="text-sm text-warm-text-muted">Started {job.date}</p>
-                 </div>
-                 <span className={`px-3 py-1 rounded-full text-xs font-medium ${job.status === 'Running' ? 'bg-warm-selected text-warm-text-secondary' : 'bg-warm-selected text-warm-state-success'}`}>
-                   {job.status}
-                 </span>
-              </div>
-
-              {job.status === 'Running' ? (
-                <div className="mt-4">
-                  <div className="h-2 w-full bg-warm-selected rounded-full overflow-hidden">
-                    <div className="h-full bg-warm-text-secondary w-[65%] rounded-full"></div>
-                  </div>
+        {isLoading ? (
+          <div className="flex items-center justify-center py-12">
+            <Loader2 className="animate-spin text-warm-text-muted" size={32} />
+          </div>
+        ) : models.length === 0 ? (
+          <div className="text-center py-12 text-warm-text-muted">
+            No model versions found. Complete a training task to create a model version.
+          </div>
+        ) : (
+          <div className="space-y-4">
+            {models.map(model => (
+              <div
+                key={model.version_id}
+                onClick={() => setSelectedModel(model)}
+                className={`bg-warm-card border rounded-lg p-5 shadow-sm cursor-pointer transition-colors ${
+                  selectedModel?.version_id === model.version_id
+                    ? 'border-warm-text-secondary'
+                    : 'border-warm-border hover:border-warm-divider'
+                }`}
+              >
+                <div className="flex justify-between items-start mb-2">
+                   <div>
+                     <h4 className="font-semibold text-warm-text-primary text-lg mb-1">
+                       {model.name}
+                       {model.is_active && <CheckCircle size={16} className="inline ml-2 text-warm-state-info" />}
+                     </h4>
+                     <p className="text-sm text-warm-text-muted">Trained {formatDate(model.trained_at)}</p>
+                   </div>
+                   <span className={`px-3 py-1 rounded-full text-xs font-medium ${
+                     model.is_active
+                       ? 'bg-warm-state-info/10 text-warm-state-info'
+                       : 'bg-warm-selected text-warm-state-success'
+                   }`}>
+                     {model.is_active ? 'Active' : model.status}
+                   </span>
                 </div>
-              ) : (
+
                 <div className="mt-4 flex gap-8">
                    <div>
-                     <span className="block text-xs text-warm-text-muted uppercase tracking-wide">Success</span>
-                     <span className="text-lg font-mono text-warm-text-secondary">{job.success}</span>
+                     <span className="block text-xs text-warm-text-muted uppercase tracking-wide">Documents</span>
+                     <span className="text-lg font-mono text-warm-text-secondary">{model.document_count}</span>
                    </div>
                    <div>
-                     <span className="block text-xs text-warm-text-muted uppercase tracking-wide">Performance</span>
-                     <span className="text-lg font-mono text-warm-text-secondary">{job.metrics}%</span>
+                     <span className="block text-xs text-warm-text-muted uppercase tracking-wide">mAP</span>
+                     <span className="text-lg font-mono text-warm-text-secondary">
+                       {model.metrics_mAP ? `${(model.metrics_mAP * 100).toFixed(1)}%` : 'N/A'}
+                     </span>
                    </div>
                    <div>
-                     <span className="block text-xs text-warm-text-muted uppercase tracking-wide">Completed</span>
-                     <span className="text-lg font-mono text-warm-text-secondary">100%</span>
+                     <span className="block text-xs text-warm-text-muted uppercase tracking-wide">Version</span>
+                     <span className="text-lg font-mono text-warm-text-secondary">{model.version}</span>
                    </div>
                 </div>
-              )}
-            </div>
-          ))}
-        </div>
+              </div>
+            ))}
+          </div>
+        )}
       </div>
 
       {/* Right: Model Detail */}
@@ -75,27 +110,34 @@ export const Models: React.FC = () => {
          <div className="bg-warm-card border border-warm-border rounded-lg p-6 shadow-card sticky top-8">
             <div className="flex justify-between items-center mb-6">
               <h3 className="text-xl font-bold text-warm-text-primary">Model Detail</h3>
-              <span className="text-sm font-medium text-warm-state-success">Completed</span>
+              <span className={`text-sm font-medium ${
+                selectedModel?.is_active ? 'text-warm-state-info' : 'text-warm-state-success'
+              }`}>
+                {selectedModel ? (selectedModel.is_active ? 'Active' : selectedModel.status) : '-'}
+              </span>
             </div>
 
             <div className="mb-8">
               <p className="text-sm text-warm-text-muted mb-1">Model name</p>
-              <p className="font-medium text-warm-text-primary">Invoices Q4 v2.1</p>
+              <p className="font-medium text-warm-text-primary">
+                {selectedModel ? `${selectedModel.name} (${selectedModel.version})` : 'Select a model'}
+              </p>
             </div>
 
             <div className="space-y-8">
                {/* Chart 1 */}
                <div>
-                  <h4 className="text-sm font-semibold text-warm-text-secondary mb-4">Bar Rate Metrics</h4>
+                  <h4 className="text-sm font-semibold text-warm-text-secondary mb-4">Model Comparison (mAP)</h4>
                   <div className="h-40">
                     <ResponsiveContainer width="100%" height="100%">
-                      <BarChart data={CHART_DATA}>
+                      <BarChart data={chartData}>
                         <CartesianGrid strokeDasharray="3 3" vertical={false} stroke="#E6E4E1" />
-                        <XAxis dataKey="name" hide />
+                        <XAxis dataKey="name" tick={{fontSize: 10, fill: '#6B6B6B'}} axisLine={false} tickLine={false} />
                         <YAxis hide domain={[0, 100]} />
-                        <Tooltip 
-                          cursor={{fill: '#F1F0ED'}} 
+                        <Tooltip
+                          cursor={{fill: '#F1F0ED'}}
                           contentStyle={{borderRadius: '8px', border: '1px solid #E6E4E1', boxShadow: '0 2px 5px rgba(0,0,0,0.05)'}}
+                          formatter={(value: number) => [`${value.toFixed(1)}%`, 'mAP']}
                         />
                         <Bar dataKey="value" fill="#3A3A3A" radius={[4, 4, 0, 0]} barSize={32} />
                       </BarChart>
@@ -105,14 +147,17 @@ export const Models: React.FC = () => {
 
                {/* Chart 2 */}
                <div>
-                  <h4 className="text-sm font-semibold text-warm-text-secondary mb-4">Entity Extraction Accuracy</h4>
+                  <h4 className="text-sm font-semibold text-warm-text-secondary mb-4">Performance Metrics</h4>
                   <div className="h-40">
                     <ResponsiveContainer width="100%" height="100%">
-                      <BarChart data={METRICS_DATA}>
+                      <BarChart data={metricsData}>
                         <CartesianGrid strokeDasharray="3 3" vertical={false} stroke="#E6E4E1" />
                         <XAxis dataKey="name" tick={{fontSize: 10, fill: '#6B6B6B'}} axisLine={false} tickLine={false} />
                         <YAxis hide domain={[0, 100]} />
-                        <Tooltip cursor={{fill: '#F1F0ED'}} />
+                        <Tooltip
+                          cursor={{fill: '#F1F0ED'}}
+                          formatter={(value: number) => [`${value.toFixed(1)}%`, 'Score']}
+                        />
                         <Bar dataKey="value" fill="#3A3A3A" radius={[4, 4, 0, 0]} barSize={32} />
                       </BarChart>
                     </ResponsiveContainer>
@@ -121,14 +166,43 @@ export const Models: React.FC = () => {
             </div>
 
             <div className="mt-8 space-y-3">
-              <Button className="w-full">Download Model</Button>
+              {selectedModel && !selectedModel.is_active ? (
+                <Button
+                  className="w-full"
+                  onClick={() => activateModel(selectedModel.version_id)}
+                  disabled={isActivating}
+                >
+                  {isActivating ? (
+                    <>
+                      <Loader2 size={16} className="mr-2 animate-spin" />
+                      Activating...
+                    </>
+                  ) : (
+                    <>
+                      <Power size={16} className="mr-2" />
+                      Activate for Inference
+                    </>
+                  )}
+                </Button>
+              ) : (
+                <Button className="w-full" disabled={!selectedModel}>
+                  {selectedModel?.is_active ? (
+                    <>
+                      <CheckCircle size={16} className="mr-2" />
+                      Currently Active
+                    </>
+                  ) : (
+                    'Select a Model'
+                  )}
+                </Button>
+              )}
               <div className="flex gap-3">
-                 <Button variant="secondary" className="flex-1">View Logs</Button>
-                 <Button variant="secondary" className="flex-1">Use as Base</Button>
+                 <Button variant="secondary" className="flex-1" disabled={!selectedModel}>View Logs</Button>
+                 <Button variant="secondary" className="flex-1" disabled={!selectedModel}>Use as Base</Button>
               </div>
             </div>
          </div>
       </div>
     </div>
   );
-};
\ No newline at end of file
+};
diff --git a/frontend/src/components/Training.tsx b/frontend/src/components/Training.tsx
index 39a9976..13c63eb 100644
--- a/frontend/src/components/Training.tsx
+++ b/frontend/src/components/Training.tsx
@@ -1,113 +1,482 @@
-import React, { useState } from 'react';
-import { Check, AlertCircle } from 'lucide-react';
-import { Button } from './Button';
-import { DocumentStatus } from '../types';
+import React, { useState, useMemo } from 'react'
+import { useQuery } from '@tanstack/react-query'
+import { Database, Plus, Trash2, Eye, Play, Check, Loader2, AlertCircle } from 'lucide-react'
+import { Button } from './Button'
+import { AugmentationConfig } from './AugmentationConfig'
+import { useDatasets } from '../hooks/useDatasets'
+import { useTrainingDocuments } from '../hooks/useTraining'
+import { trainingApi } from '../api/endpoints'
+import type { DatasetListItem } from '../api/types'
+import type { AugmentationConfig as AugmentationConfigType } from '../api/endpoints/augmentation'
 
-export const Training: React.FC = () => {
-  const [split, setSplit] = useState(80);
+type Tab = 'datasets' | 'create'
 
-  const docs = [
-    { id: '1', name: 'Document Document 1', date: '12/28/2024', status: DocumentStatus.VERIFIED },
-    { id: '2', name: 'Document Document 2', date: '12/29/2024', status: DocumentStatus.VERIFIED },
-    { id: '3', name: 'Document Document 3', date: '12/29/2024', status: DocumentStatus.VERIFIED },
-    { id: '4', name: 'Document Document 4', date: '12/29/2024', status: DocumentStatus.PARTIAL },
-    { id: '5', name: 'Document Document 5', date: '12/29/2024', status: DocumentStatus.PARTIAL },
-    { id: '6', name: 'Document Document 6', date: '12/29/2024', status: DocumentStatus.PARTIAL },
-    { id: '8', name: 'Document Document 8', date: '12/29/2024', status: DocumentStatus.VERIFIED },
-  ];
+interface TrainingProps {
+  onNavigate?: (view: string, id?: string) => void
+}
+
+const STATUS_STYLES: Record<string, string> = {
+  ready: 'bg-warm-state-success/10 text-warm-state-success',
+  building: 'bg-warm-state-info/10 text-warm-state-info',
+  training: 'bg-warm-state-info/10 text-warm-state-info',
+  failed: 'bg-warm-state-error/10 text-warm-state-error',
+  pending: 'bg-warm-state-warning/10 text-warm-state-warning',
+  scheduled: 'bg-warm-state-warning/10 text-warm-state-warning',
+  running: 'bg-warm-state-info/10 text-warm-state-info',
+}
+
+const StatusBadge: React.FC<{ status: string; trainingStatus?: string | null }> = ({ status, trainingStatus }) => {
+  // If there's an active training task, show training status
+  const displayStatus = trainingStatus === 'running'
+    ? 'training'
+    : trainingStatus === 'pending' || trainingStatus === 'scheduled'
+      ? 'pending'
+      : status
 
   return (
-    <div className="p-8 max-w-7xl mx-auto h-[calc(100vh-56px)] flex gap-8">
-      {/* Document Selection List */}
-      <div className="flex-1 flex flex-col">
-        <h2 className="text-2xl font-bold text-warm-text-primary mb-6">Document Selection</h2>
-        
-        <div className="flex-1 bg-warm-card border border-warm-border rounded-lg overflow-hidden flex flex-col shadow-sm">
-           <div className="overflow-auto flex-1">
-             <table className="w-full text-left">
-              <thead className="sticky top-0 bg-white border-b border-warm-border z-10">
-                <tr>
-                   <th className="py-3 pl-6 pr-4 w-12"><input type="checkbox" className="rounded border-warm-divider"/></th>
-                   <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase">Document name</th>
-                   <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase">Date</th>
-                   <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase">Status</th>
-                </tr>
-              </thead>
-              <tbody>
-                {docs.map(doc => (
-                  <tr key={doc.id} className="border-b border-warm-border hover:bg-warm-hover transition-colors">
-                     <td className="py-3 pl-6 pr-4"><input type="checkbox" defaultChecked className="rounded border-warm-divider accent-warm-state-info"/></td>
-                     <td className="py-3 px-4 text-sm font-medium text-warm-text-secondary">{doc.name}</td>
-                     <td className="py-3 px-4 text-sm text-warm-text-muted font-mono">{doc.date}</td>
-                     <td className="py-3 px-4">
-                       {doc.status === DocumentStatus.VERIFIED ? (
-                         <div className="flex items-center text-warm-state-success text-sm font-medium">
-                            <div className="w-5 h-5 rounded-full bg-warm-state-success flex items-center justify-center text-white mr-2">
-                              <Check size={12} strokeWidth={3}/>
-                            </div>
-                            Verified
-                         </div>
-                       ) : (
-                         <div className="flex items-center text-warm-text-muted text-sm">
-                            <div className="w-5 h-5 rounded-full bg-[#BDBBB5] flex items-center justify-center text-white mr-2">
-                              <span className="font-bold text-[10px]">!</span>
-                            </div>
-                            Partial
-                         </div>
-                       )}
-                     </td>
-                  </tr>
-                ))}
-              </tbody>
-             </table>
-           </div>
-        </div>
-      </div>
+    <span className={`inline-flex items-center px-2.5 py-1 rounded-full text-xs font-medium ${STATUS_STYLES[displayStatus] ?? 'bg-warm-border text-warm-text-muted'}`}>
+      {(displayStatus === 'building' || displayStatus === 'training') && <Loader2 size={12} className="mr-1 animate-spin" />}
+      {displayStatus === 'ready' && <Check size={12} className="mr-1" />}
+      {displayStatus === 'failed' && <AlertCircle size={12} className="mr-1" />}
+      {displayStatus}
+    </span>
+  )
+}
 
-      {/* Configuration Panel */}
-      <div className="w-96">
-        <div className="bg-warm-card rounded-lg border border-warm-border shadow-card p-6 sticky top-8">
-          <h3 className="text-lg font-semibold text-warm-text-primary mb-6">Training Configuration</h3>
-          
-          <div className="space-y-6">
-            <div>
-              <label className="block text-sm font-medium text-warm-text-secondary mb-2">Model Name</label>
-              <input 
-                type="text" 
-                placeholder="e.g. Invoices Q4"
+// --- Train Dialog ---
+
+interface TrainDialogProps {
+  dataset: DatasetListItem
+  onClose: () => void
+  onSubmit: (config: {
+    name: string
+    config: {
+      model_name?: string
+      base_model_version_id?: string | null
+      epochs: number
+      batch_size: number
+      augmentation?: AugmentationConfigType
+      augmentation_multiplier?: number
+    }
+  }) => void
+  isPending: boolean
+}
+
+const TrainDialog: React.FC<TrainDialogProps> = ({ dataset, onClose, onSubmit, isPending }) => {
+  const [name, setName] = useState(`train-${dataset.name}`)
+  const [epochs, setEpochs] = useState(100)
+  const [batchSize, setBatchSize] = useState(16)
+  const [baseModelType, setBaseModelType] = useState<'pretrained' | 'existing'>('pretrained')
+  const [baseModelVersionId, setBaseModelVersionId] = useState<string | null>(null)
+  const [augmentationEnabled, setAugmentationEnabled] = useState(false)
+  const [augmentationConfig, setAugmentationConfig] = useState<Partial<AugmentationConfigType>>({})
+  const [augmentationMultiplier, setAugmentationMultiplier] = useState(2)
+
+  // Fetch available trained models
+  const { data: modelsData } = useQuery({
+    queryKey: ['training', 'models', 'completed'],
+    queryFn: () => trainingApi.getModels({ status: 'completed' }),
+  })
+  const completedModels = modelsData?.models ?? []
+
+  const handleSubmit = () => {
+    onSubmit({
+      name,
+      config: {
+        model_name: baseModelType === 'pretrained' ? 'yolo11n.pt' : undefined,
+        base_model_version_id: baseModelType === 'existing' ? baseModelVersionId : null,
+        epochs,
+        batch_size: batchSize,
+        augmentation: augmentationEnabled
+          ? (augmentationConfig as AugmentationConfigType)
+          : undefined,
+        augmentation_multiplier: augmentationEnabled ? augmentationMultiplier : undefined,
+      },
+    })
+  }
+
+  return (
+    <div className="fixed inset-0 bg-black/40 flex items-center justify-center z-50" onClick={onClose}>
+      <div className="bg-white rounded-lg border border-warm-border shadow-lg w-[480px] max-h-[90vh] overflow-y-auto p-6" onClick={e => e.stopPropagation()}>
+        <h3 className="text-lg font-semibold text-warm-text-primary mb-4">Start Training</h3>
+        <p className="text-sm text-warm-text-muted mb-4">
+          Dataset: <span className="font-medium text-warm-text-secondary">{dataset.name}</span>
+          {' '}({dataset.total_images} images, {dataset.total_annotations} annotations)
+        </p>
+
+        <div className="space-y-4">
+          <div>
+            <label className="block text-sm font-medium text-warm-text-secondary mb-1">Task Name</label>
+            <input type="text" value={name} onChange={e => setName(e.target.value)}
+              className="w-full h-10 px-3 rounded-md border border-warm-divider bg-white text-warm-text-primary focus:outline-none focus:ring-1 focus:ring-warm-state-info" />
+          </div>
+
+          {/* Base Model Selection */}
+          <div>
+            <label className="block text-sm font-medium text-warm-text-secondary mb-1">Base Model</label>
+            <select
+              value={baseModelType === 'pretrained' ? 'pretrained' : baseModelVersionId ?? ''}
+              onChange={e => {
+                if (e.target.value === 'pretrained') {
+                  setBaseModelType('pretrained')
+                  setBaseModelVersionId(null)
+                } else {
+                  setBaseModelType('existing')
+                  setBaseModelVersionId(e.target.value)
+                }
+              }}
+              className="w-full h-10 px-3 rounded-md border border-warm-divider bg-white text-warm-text-primary focus:outline-none focus:ring-1 focus:ring-warm-state-info"
+            >
+              <option value="pretrained">yolo11n.pt (Pretrained)</option>
+              {completedModels.map(m => (
+                <option key={m.task_id} value={m.task_id}>
+                  {m.name} ({m.metrics_mAP ? `${(m.metrics_mAP * 100).toFixed(1)}% mAP` : 'No metrics'})
+                </option>
+              ))}
+            </select>
+            <p className="text-xs text-warm-text-muted mt-1">
+              {baseModelType === 'pretrained'
+                ? 'Start from pretrained YOLO model'
+                : 'Continue training from an existing model (incremental training)'}
+            </p>
+          </div>
+
+          <div className="flex gap-4">
+            <div className="flex-1">
+              <label htmlFor="train-epochs" className="block text-sm font-medium text-warm-text-secondary mb-1">Epochs</label>
+              <input
+                id="train-epochs"
+                type="number"
+                min={1}
+                max={1000}
+                value={epochs}
+                onChange={e => setEpochs(Math.max(1, Math.min(1000, Number(e.target.value) || 1)))}
                 className="w-full h-10 px-3 rounded-md border border-warm-divider bg-white text-warm-text-primary focus:outline-none focus:ring-1 focus:ring-warm-state-info"
               />
             </div>
-
-            <div>
-              <label className="block text-sm font-medium text-warm-text-secondary mb-2">Base Model</label>
-              <select className="w-full h-10 px-3 rounded-md border border-warm-divider bg-white text-warm-text-primary focus:outline-none focus:ring-1 focus:ring-warm-state-info appearance-none">
-                 <option>LayoutLMv3 (Standard)</option>
-                 <option>Donut (Beta)</option>
-              </select>
-            </div>
-
-            <div>
-              <div className="flex justify-between mb-2">
-                <label className="block text-sm font-medium text-warm-text-secondary">Train/Test Split</label>
-                <span className="text-xs font-mono text-warm-text-muted">{split}% / {100-split}%</span>
-              </div>
-              <input 
-                type="range" 
-                min="50" 
-                max="95" 
-                value={split} 
-                onChange={(e) => setSplit(parseInt(e.target.value))}
-                className="w-full h-1.5 bg-warm-border rounded-lg appearance-none cursor-pointer accent-warm-state-info"
+            <div className="flex-1">
+              <label htmlFor="train-batch-size" className="block text-sm font-medium text-warm-text-secondary mb-1">Batch Size</label>
+              <input
+                id="train-batch-size"
+                type="number"
+                min={1}
+                max={128}
+                value={batchSize}
+                onChange={e => setBatchSize(Math.max(1, Math.min(128, Number(e.target.value) || 1)))}
+                className="w-full h-10 px-3 rounded-md border border-warm-divider bg-white text-warm-text-primary focus:outline-none focus:ring-1 focus:ring-warm-state-info"
               />
             </div>
+          </div>
+
+          {/* Augmentation Configuration */}
+          <AugmentationConfig
+            enabled={augmentationEnabled}
+            onEnabledChange={setAugmentationEnabled}
+            config={augmentationConfig}
+            onConfigChange={setAugmentationConfig}
+          />
+
+          {/* Augmentation Multiplier - only shown when augmentation is enabled */}
+          {augmentationEnabled && (
+            <div>
+              <label htmlFor="aug-multiplier" className="block text-sm font-medium text-warm-text-secondary mb-1">
+                Augmentation Multiplier
+              </label>
+              <input
+                id="aug-multiplier"
+                type="number"
+                min={1}
+                max={10}
+                value={augmentationMultiplier}
+                onChange={e => setAugmentationMultiplier(Math.max(1, Math.min(10, Number(e.target.value) || 1)))}
+                className="w-full h-10 px-3 rounded-md border border-warm-divider bg-white text-warm-text-primary focus:outline-none focus:ring-1 focus:ring-warm-state-info"
+              />
+              <p className="text-xs text-warm-text-muted mt-1">
+                Number of augmented copies per original image (1-10)
+              </p>
+            </div>
+          )}
+        </div>
+
+        <div className="flex justify-end gap-3 mt-6">
+          <Button variant="secondary" onClick={onClose} disabled={isPending}>Cancel</Button>
+          <Button onClick={handleSubmit} disabled={isPending || !name.trim()}>
+            {isPending ? <><Loader2 size={14} className="mr-1 animate-spin" />Training...</> : 'Start Training'}
+          </Button>
+        </div>
+      </div>
+    </div>
+  )
+}
+
+// --- Dataset List ---
+
+const DatasetList: React.FC<{
+  onNavigate?: (view: string, id?: string) => void
+  onSwitchTab: (tab: Tab) => void
+}> = ({ onNavigate, onSwitchTab }) => {
+  const { datasets, isLoading, deleteDataset, isDeleting, trainFromDataset, isTraining } = useDatasets()
+  const [trainTarget, setTrainTarget] = useState<DatasetListItem | null>(null)
+
+  const handleTrain = (config: {
+    name: string
+    config: {
+      model_name?: string
+      base_model_version_id?: string | null
+      epochs: number
+      batch_size: number
+      augmentation?: AugmentationConfigType
+      augmentation_multiplier?: number
+    }
+  }) => {
+    if (!trainTarget) return
+    // Pass config to the training API
+    const trainRequest = {
+      name: config.name,
+      config: config.config,
+    }
+    trainFromDataset(
+      { datasetId: trainTarget.dataset_id, req: trainRequest },
+      { onSuccess: () => setTrainTarget(null) },
+    )
+  }
+
+  if (isLoading) {
+    return <div className="flex items-center justify-center py-20 text-warm-text-muted"><Loader2 size={24} className="animate-spin mr-2" />Loading datasets...</div>
+  }
+
+  if (datasets.length === 0) {
+    return (
+      <div className="flex flex-col items-center justify-center py-20 text-warm-text-muted">
+        <Database size={48} className="mb-4 opacity-40" />
+        <p className="text-lg mb-2">No datasets yet</p>
+        <p className="text-sm mb-4">Create a dataset to start training</p>
+        <Button onClick={() => onSwitchTab('create')}><Plus size={14} className="mr-1" />Create Dataset</Button>
+      </div>
+    )
+  }
+
+  return (
+    <>
+      <div className="bg-warm-card border border-warm-border rounded-lg overflow-hidden shadow-sm">
+        <table className="w-full text-left">
+          <thead className="bg-white border-b border-warm-border">
+            <tr>
+              <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase">Name</th>
+              <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase">Status</th>
+              <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase">Docs</th>
+              <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase">Images</th>
+              <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase">Annotations</th>
+              <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase">Created</th>
+              <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase">Actions</th>
+            </tr>
+          </thead>
+          <tbody>
+            {datasets.map(ds => (
+              <tr key={ds.dataset_id} className="border-b border-warm-border hover:bg-warm-hover transition-colors">
+                <td className="py-3 px-4 text-sm font-medium text-warm-text-secondary">{ds.name}</td>
+                <td className="py-3 px-4"><StatusBadge status={ds.status} trainingStatus={ds.training_status} /></td>
+                <td className="py-3 px-4 text-sm text-warm-text-muted font-mono">{ds.total_documents}</td>
+                <td className="py-3 px-4 text-sm text-warm-text-muted font-mono">{ds.total_images}</td>
+                <td className="py-3 px-4 text-sm text-warm-text-muted font-mono">{ds.total_annotations}</td>
+                <td className="py-3 px-4 text-sm text-warm-text-muted">{new Date(ds.created_at).toLocaleDateString()}</td>
+                <td className="py-3 px-4">
+                  <div className="flex gap-1">
+                    <button title="View" onClick={() => onNavigate?.('dataset-detail', ds.dataset_id)}
+                      className="p-1.5 rounded hover:bg-warm-selected text-warm-text-muted hover:text-warm-state-info transition-colors">
+                      <Eye size={14} />
+                    </button>
+                    {ds.status === 'ready' && (
+                      <button title="Train" onClick={() => setTrainTarget(ds)}
+                        className="p-1.5 rounded hover:bg-warm-selected text-warm-text-muted hover:text-warm-state-success transition-colors">
+                        <Play size={14} />
+                      </button>
+                    )}
+                    <button title="Delete" onClick={() => deleteDataset(ds.dataset_id)}
+                      disabled={isDeleting}
+                      className="p-1.5 rounded hover:bg-warm-selected text-warm-text-muted hover:text-warm-state-error transition-colors">
+                      <Trash2 size={14} />
+                    </button>
+                  </div>
+                </td>
+              </tr>
+            ))}
+          </tbody>
+        </table>
+      </div>
+
+      {trainTarget && (
+        <TrainDialog dataset={trainTarget} onClose={() => setTrainTarget(null)} onSubmit={handleTrain} isPending={isTraining} />
+      )}
+    </>
+  )
+}
+
+// --- Create Dataset ---
+
+const CreateDataset: React.FC<{ onSwitchTab: (tab: Tab) => void }> = ({ onSwitchTab }) => {
+  const { documents, isLoading: isLoadingDocs } = useTrainingDocuments({ has_annotations: true })
+  const { createDatasetAsync, isCreating } = useDatasets()
+
+  const [selectedIds, setSelectedIds] = useState<Set<string>>(new Set())
+  const [name, setName] = useState('')
+  const [description, setDescription] = useState('')
+  const [trainRatio, setTrainRatio] = useState(0.7)
+  const [valRatio, setValRatio] = useState(0.2)
+
+  const testRatio = useMemo(() => Math.max(0, +(1 - trainRatio - valRatio).toFixed(2)), [trainRatio, valRatio])
+
+  const toggleDoc = (id: string) => {
+    setSelectedIds(prev => {
+      const next = new Set(prev)
+      if (next.has(id)) { next.delete(id) } else { next.add(id) }
+      return next
+    })
+  }
+
+  const toggleAll = () => {
+    if (selectedIds.size === documents.length) {
+      setSelectedIds(new Set())
+    } else {
+      setSelectedIds(new Set(documents.map((d) => d.document_id)))
+    }
+  }
+
+  const handleCreate = async () => {
+    await createDatasetAsync({
+      name,
+      description: description || undefined,
+      document_ids: [...selectedIds],
+      train_ratio: trainRatio,
+      val_ratio: valRatio,
+    })
+    onSwitchTab('datasets')
+  }
+
+  return (
+    <div className="flex gap-8">
+      {/* Document selection */}
+      <div className="flex-1 flex flex-col">
+        <h3 className="text-lg font-semibold text-warm-text-primary mb-4">Select Documents</h3>
+        {isLoadingDocs ? (
+          <div className="flex items-center justify-center py-12 text-warm-text-muted"><Loader2 size={20} className="animate-spin mr-2" />Loading...</div>
+        ) : (
+          <div className="bg-warm-card border border-warm-border rounded-lg overflow-hidden shadow-sm flex-1">
+            <div className="overflow-auto max-h-[calc(100vh-240px)]">
+              <table className="w-full text-left">
+                <thead className="sticky top-0 bg-white border-b border-warm-border z-10">
+                  <tr>
+                    <th className="py-3 pl-6 pr-4 w-12">
+                      <input type="checkbox" checked={selectedIds.size === documents.length && documents.length > 0}
+                        onChange={toggleAll} className="rounded border-warm-divider accent-warm-state-info" />
+                    </th>
+                    <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase">Document ID</th>
+                    <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase">Pages</th>
+                    <th className="py-3 px-4 text-xs font-semibold text-warm-text-muted uppercase">Annotations</th>
+                  </tr>
+                </thead>
+                <tbody>
+                  {documents.map((doc) => (
+                    <tr key={doc.document_id} className="border-b border-warm-border hover:bg-warm-hover transition-colors cursor-pointer"
+                      onClick={() => toggleDoc(doc.document_id)}>
+                      <td className="py-3 pl-6 pr-4">
+                        <input type="checkbox" checked={selectedIds.has(doc.document_id)} readOnly
+                          className="rounded border-warm-divider accent-warm-state-info pointer-events-none" />
+                      </td>
+                      <td className="py-3 px-4 text-sm font-mono text-warm-text-secondary">{doc.document_id.slice(0, 8)}...</td>
+                      <td className="py-3 px-4 text-sm text-warm-text-muted font-mono">{doc.page_count}</td>
+                      <td className="py-3 px-4 text-sm text-warm-text-muted font-mono">{doc.annotation_count ?? 0}</td>
+                    </tr>
+                  ))}
+                </tbody>
+              </table>
+            </div>
+          </div>
+        )}
+        <p className="text-sm text-warm-text-muted mt-2">{selectedIds.size} of {documents.length} documents selected</p>
+      </div>
+
+      {/* Config panel */}
+      <div className="w-80">
+        <div className="bg-warm-card rounded-lg border border-warm-border shadow-card p-6 sticky top-8">
+          <h3 className="text-lg font-semibold text-warm-text-primary mb-4">Dataset Configuration</h3>
+          <div className="space-y-4">
+            <div>
+              <label className="block text-sm font-medium text-warm-text-secondary mb-1">Name</label>
+              <input type="text" value={name} onChange={e => setName(e.target.value)} placeholder="e.g. invoice-dataset-v1"
+                className="w-full h-10 px-3 rounded-md border border-warm-divider bg-white text-warm-text-primary focus:outline-none focus:ring-1 focus:ring-warm-state-info" />
+            </div>
+            <div>
+              <label className="block text-sm font-medium text-warm-text-secondary mb-1">Description</label>
+              <textarea value={description} onChange={e => setDescription(e.target.value)} rows={2} placeholder="Optional"
+                className="w-full px-3 py-2 rounded-md border border-warm-divider bg-white text-warm-text-primary focus:outline-none focus:ring-1 focus:ring-warm-state-info resize-none" />
+            </div>
+            <div>
+              <label className="block text-sm font-medium text-warm-text-secondary mb-1">Train / Val / Test Split</label>
+              <div className="flex gap-2 text-sm">
+                <div className="flex-1">
+                  <span className="text-xs text-warm-text-muted">Train</span>
+                  <input type="number" step={0.05} min={0.1} max={0.9} value={trainRatio} onChange={e => setTrainRatio(Number(e.target.value))}
+                    className="w-full h-9 px-2 rounded-md border border-warm-divider bg-white text-warm-text-primary text-center font-mono focus:outline-none focus:ring-1 focus:ring-warm-state-info" />
+                </div>
+                <div className="flex-1">
+                  <span className="text-xs text-warm-text-muted">Val</span>
+                  <input type="number" step={0.05} min={0} max={0.5} value={valRatio} onChange={e => setValRatio(Number(e.target.value))}
+                    className="w-full h-9 px-2 rounded-md border border-warm-divider bg-white text-warm-text-primary text-center font-mono focus:outline-none focus:ring-1 focus:ring-warm-state-info" />
+                </div>
+                <div className="flex-1">
+                  <span className="text-xs text-warm-text-muted">Test</span>
+                  <input type="number" value={testRatio} readOnly
+                    className="w-full h-9 px-2 rounded-md border border-warm-divider bg-warm-hover text-warm-text-muted text-center font-mono" />
+                </div>
+              </div>
+            </div>
 
             <div className="pt-4 border-t border-warm-border">
-              <Button className="w-full h-12">Start Training</Button>
+              {selectedIds.size > 0 && selectedIds.size < 10 && (
+                <p className="text-xs text-warm-state-warning mb-2">
+                  Minimum 10 documents required for training ({selectedIds.size}/10 selected)
+                </p>
+              )}
+              <Button className="w-full h-11" onClick={handleCreate}
+                disabled={isCreating || selectedIds.size < 10 || !name.trim()}>
+                {isCreating ? <><Loader2 size={14} className="mr-1 animate-spin" />Creating...</> : <><Plus size={14} className="mr-1" />Create Dataset</>}
+              </Button>
             </div>
           </div>
         </div>
       </div>
     </div>
-  );
-};
\ No newline at end of file
+  )
+}
+
+// --- Main Training Component ---
+
+export const Training: React.FC<TrainingProps> = ({ onNavigate }) => {
+  const [activeTab, setActiveTab] = useState<Tab>('datasets')
+
+  return (
+    <div className="p-8 max-w-7xl mx-auto">
+      <div className="flex items-center justify-between mb-6">
+        <h2 className="text-2xl font-bold text-warm-text-primary">Training</h2>
+      </div>
+
+      {/* Tabs */}
+      <div className="flex gap-1 mb-6 border-b border-warm-border">
+        {([['datasets', 'Datasets'], ['create', 'Create Dataset']] as const).map(([key, label]) => (
+          <button key={key} onClick={() => setActiveTab(key)}
+            className={`px-4 py-2.5 text-sm font-medium border-b-2 transition-colors ${
+              activeTab === key
+                ? 'border-warm-state-info text-warm-state-info'
+                : 'border-transparent text-warm-text-muted hover:text-warm-text-secondary'
+            }`}>
+            {label}
+          </button>
+        ))}
+      </div>
+
+      {activeTab === 'datasets' && <DatasetList onNavigate={onNavigate} onSwitchTab={setActiveTab} />}
+      {activeTab === 'create' && <CreateDataset onSwitchTab={setActiveTab} />}
+    </div>
+  )
+}
diff --git a/frontend/src/components/UploadModal.tsx b/frontend/src/components/UploadModal.tsx
index ca7b9b3..f76df93 100644
--- a/frontend/src/components/UploadModal.tsx
+++ b/frontend/src/components/UploadModal.tsx
@@ -11,6 +11,7 @@ interface UploadModalProps {
 export const UploadModal: React.FC<UploadModalProps> = ({ isOpen, onClose }) => {
   const [isDragging, setIsDragging] = useState(false)
   const [selectedFiles, setSelectedFiles] = useState<File[]>([])
+  const [groupKey, setGroupKey] = useState('')
   const [uploadStatus, setUploadStatus] = useState<'idle' | 'uploading' | 'success' | 'error'>('idle')
   const [errorMessage, setErrorMessage] = useState('')
   const fileInputRef = useRef<HTMLInputElement>(null)
@@ -61,10 +62,13 @@ export const UploadModal: React.FC<UploadModalProps> = ({ isOpen, onClose }) =>
       // Upload files one by one
       for (const file of selectedFiles) {
         await new Promise<void>((resolve, reject) => {
-          uploadDocument(file, {
-            onSuccess: () => resolve(),
-            onError: (error: Error) => reject(error),
-          })
+          uploadDocument(
+            { file, groupKey: groupKey || undefined },
+            {
+              onSuccess: () => resolve(),
+              onError: (error: Error) => reject(error),
+            }
+          )
         })
       }
 
@@ -72,6 +76,7 @@ export const UploadModal: React.FC<UploadModalProps> = ({ isOpen, onClose }) =>
       setTimeout(() => {
         onClose()
         setSelectedFiles([])
+        setGroupKey('')
         setUploadStatus('idle')
       }, 1500)
     } catch (error) {
@@ -85,6 +90,7 @@ export const UploadModal: React.FC<UploadModalProps> = ({ isOpen, onClose }) =>
       return // Prevent closing during upload
     }
     setSelectedFiles([])
+    setGroupKey('')
     setUploadStatus('idle')
     setErrorMessage('')
     onClose()
@@ -173,6 +179,26 @@ export const UploadModal: React.FC<UploadModalProps> = ({ isOpen, onClose }) =>
           </div>
         )}
 
+        {/* Group Key Input */}
+        {selectedFiles.length > 0 && (
+          <div className="mb-6">
+            <label className="block text-sm font-medium text-warm-text-secondary mb-2">
+              Group Key (optional)
+            </label>
+            <input
+              type="text"
+              value={groupKey}
+              onChange={(e) => setGroupKey(e.target.value)}
+              placeholder="e.g., 2024-Q1, supplier-abc, project-name"
+              className="w-full px-3 h-10 rounded-md border border-warm-border bg-white text-sm text-warm-text-secondary focus:outline-none focus:ring-1 focus:ring-warm-state-info transition-shadow"
+              disabled={uploadStatus === 'uploading'}
+            />
+            <p className="text-xs text-warm-text-muted mt-1">
+              Use group keys to organize documents into logical groups
+            </p>
+          </div>
+        )}
+
         {/* Status Messages */}
         {uploadStatus === 'success' && (
           <div className="mb-4 p-3 bg-green-50 border border-green-200 rounded flex items-center gap-2">
diff --git a/frontend/src/hooks/index.ts b/frontend/src/hooks/index.ts
index d642394..d1b5a7b 100644
--- a/frontend/src/hooks/index.ts
+++ b/frontend/src/hooks/index.ts
@@ -2,3 +2,6 @@ export { useDocuments } from './useDocuments'
 export { useDocumentDetail } from './useDocumentDetail'
 export { useAnnotations } from './useAnnotations'
 export { useTraining, useTrainingDocuments } from './useTraining'
+export { useDatasets, useDatasetDetail } from './useDatasets'
+export { useAugmentation } from './useAugmentation'
+export { useModels, useModelDetail, useActiveModel } from './useModels'
diff --git a/frontend/src/hooks/useAugmentation.test.tsx b/frontend/src/hooks/useAugmentation.test.tsx
new file mode 100644
index 0000000..33a5dde
--- /dev/null
+++ b/frontend/src/hooks/useAugmentation.test.tsx
@@ -0,0 +1,226 @@
+/**
+ * Tests for useAugmentation hook.
+ *
+ * TDD Phase 1: RED - Write tests first, then implement to pass.
+ */
+
+import { describe, it, expect, vi, beforeEach } from 'vitest'
+import { renderHook, waitFor } from '@testing-library/react'
+import { QueryClient, QueryClientProvider } from '@tanstack/react-query'
+import { augmentationApi } from '../api/endpoints/augmentation'
+import { useAugmentation } from './useAugmentation'
+import type { ReactNode } from 'react'
+
+// Mock the API
+vi.mock('../api/endpoints/augmentation', () => ({
+  augmentationApi: {
+    getTypes: vi.fn(),
+    getPresets: vi.fn(),
+    preview: vi.fn(),
+    previewConfig: vi.fn(),
+    createBatch: vi.fn(),
+  },
+}))
+
+// Test wrapper with QueryClient
+const createWrapper = () => {
+  const queryClient = new QueryClient({
+    defaultOptions: {
+      queries: {
+        retry: false,
+      },
+    },
+  })
+  return ({ children }: { children: ReactNode }) => (
+    <QueryClientProvider client={queryClient}>{children}</QueryClientProvider>
+  )
+}
+
+describe('useAugmentation', () => {
+  beforeEach(() => {
+    vi.clearAllMocks()
+  })
+
+  describe('getTypes', () => {
+    it('should fetch augmentation types', async () => {
+      const mockTypes = {
+        augmentation_types: [
+          {
+            name: 'gaussian_noise',
+            description: 'Adds Gaussian noise',
+            affects_geometry: false,
+            stage: 'noise',
+            default_params: { mean: 0, std: 15 },
+          },
+          {
+            name: 'perspective_warp',
+            description: 'Applies perspective warp',
+            affects_geometry: true,
+            stage: 'geometric',
+            default_params: { max_warp: 0.02 },
+          },
+        ],
+      }
+      vi.mocked(augmentationApi.getTypes).mockResolvedValueOnce(mockTypes)
+
+      const { result } = renderHook(() => useAugmentation(), {
+        wrapper: createWrapper(),
+      })
+
+      await waitFor(() => {
+        expect(result.current.isLoadingTypes).toBe(false)
+      })
+
+      expect(result.current.augmentationTypes).toHaveLength(2)
+      expect(result.current.augmentationTypes[0].name).toBe('gaussian_noise')
+    })
+
+    it('should handle error when fetching types', async () => {
+      vi.mocked(augmentationApi.getTypes).mockRejectedValueOnce(new Error('Network error'))
+
+      const { result } = renderHook(() => useAugmentation(), {
+        wrapper: createWrapper(),
+      })
+
+      await waitFor(() => {
+        expect(result.current.isLoadingTypes).toBe(false)
+      })
+
+      expect(result.current.typesError).toBeTruthy()
+    })
+  })
+
+  describe('getPresets', () => {
+    it('should fetch augmentation presets', async () => {
+      const mockPresets = {
+        presets: [
+          { name: 'conservative', description: 'Safe augmentations' },
+          { name: 'moderate', description: 'Balanced augmentations' },
+          { name: 'aggressive', description: 'Strong augmentations' },
+        ],
+      }
+      vi.mocked(augmentationApi.getTypes).mockResolvedValueOnce({ augmentation_types: [] })
+      vi.mocked(augmentationApi.getPresets).mockResolvedValueOnce(mockPresets)
+
+      const { result } = renderHook(() => useAugmentation(), {
+        wrapper: createWrapper(),
+      })
+
+      await waitFor(() => {
+        expect(result.current.isLoadingPresets).toBe(false)
+      })
+
+      expect(result.current.presets).toHaveLength(3)
+      expect(result.current.presets[0].name).toBe('conservative')
+    })
+  })
+
+  describe('preview', () => {
+    it('should preview single augmentation', async () => {
+      const mockPreview = {
+        preview_url: 'data:image/png;base64,xxx',
+        original_url: 'data:image/png;base64,yyy',
+        applied_params: { std: 15 },
+      }
+      vi.mocked(augmentationApi.getTypes).mockResolvedValueOnce({ augmentation_types: [] })
+      vi.mocked(augmentationApi.getPresets).mockResolvedValueOnce({ presets: [] })
+      vi.mocked(augmentationApi.preview).mockResolvedValueOnce(mockPreview)
+
+      const { result } = renderHook(() => useAugmentation(), {
+        wrapper: createWrapper(),
+      })
+
+      await waitFor(() => {
+        expect(result.current.isLoadingTypes).toBe(false)
+      })
+
+      // Call preview mutation
+      result.current.preview({
+        documentId: 'doc-123',
+        augmentationType: 'gaussian_noise',
+        params: { std: 15 },
+        page: 1,
+      })
+
+      await waitFor(() => {
+        expect(augmentationApi.preview).toHaveBeenCalledWith(
+          'doc-123',
+          { augmentation_type: 'gaussian_noise', params: { std: 15 } },
+          1
+        )
+      })
+    })
+
+    it('should track preview loading state', async () => {
+      vi.mocked(augmentationApi.getTypes).mockResolvedValueOnce({ augmentation_types: [] })
+      vi.mocked(augmentationApi.getPresets).mockResolvedValueOnce({ presets: [] })
+      vi.mocked(augmentationApi.preview).mockImplementation(
+        () => new Promise((resolve) => setTimeout(resolve, 100))
+      )
+
+      const { result } = renderHook(() => useAugmentation(), {
+        wrapper: createWrapper(),
+      })
+
+      await waitFor(() => {
+        expect(result.current.isLoadingTypes).toBe(false)
+      })
+
+      expect(result.current.isPreviewing).toBe(false)
+
+      result.current.preview({
+        documentId: 'doc-123',
+        augmentationType: 'gaussian_noise',
+        params: {},
+        page: 1,
+      })
+
+      // State update happens asynchronously
+      await waitFor(() => {
+        expect(result.current.isPreviewing).toBe(true)
+      })
+    })
+  })
+
+  describe('createBatch', () => {
+    it('should create augmented dataset', async () => {
+      const mockResponse = {
+        task_id: 'task-123',
+        status: 'pending',
+        message: 'Augmentation task queued',
+        estimated_images: 100,
+      }
+      vi.mocked(augmentationApi.getTypes).mockResolvedValueOnce({ augmentation_types: [] })
+      vi.mocked(augmentationApi.getPresets).mockResolvedValueOnce({ presets: [] })
+      vi.mocked(augmentationApi.createBatch).mockResolvedValueOnce(mockResponse)
+
+      const { result } = renderHook(() => useAugmentation(), {
+        wrapper: createWrapper(),
+      })
+
+      await waitFor(() => {
+        expect(result.current.isLoadingTypes).toBe(false)
+      })
+
+      result.current.createBatch({
+        dataset_id: 'dataset-123',
+        config: {
+          gaussian_noise: { enabled: true, probability: 0.5, params: {} },
+        },
+        output_name: 'augmented-dataset',
+        multiplier: 2,
+      })
+
+      await waitFor(() => {
+        expect(augmentationApi.createBatch).toHaveBeenCalledWith({
+          dataset_id: 'dataset-123',
+          config: {
+            gaussian_noise: { enabled: true, probability: 0.5, params: {} },
+          },
+          output_name: 'augmented-dataset',
+          multiplier: 2,
+        })
+      })
+    })
+  })
+})
diff --git a/frontend/src/hooks/useAugmentation.ts b/frontend/src/hooks/useAugmentation.ts
new file mode 100644
index 0000000..dd299ce
--- /dev/null
+++ b/frontend/src/hooks/useAugmentation.ts
@@ -0,0 +1,121 @@
+/**
+ * Hook for managing augmentation operations.
+ *
+ * Provides functions for fetching augmentation types, presets, and previewing augmentations.
+ */
+
+import { useQuery, useMutation } from '@tanstack/react-query'
+import {
+  augmentationApi,
+  type AugmentationTypesResponse,
+  type PresetsResponse,
+  type PreviewResponse,
+  type BatchRequest,
+  type BatchResponse,
+  type AugmentationConfig,
+} from '../api/endpoints/augmentation'
+
+interface PreviewParams {
+  documentId: string
+  augmentationType: string
+  params: Record<string, unknown>
+  page?: number
+}
+
+interface PreviewConfigParams {
+  documentId: string
+  config: AugmentationConfig
+  page?: number
+}
+
+export const useAugmentation = () => {
+  // Fetch augmentation types
+  const {
+    data: typesData,
+    isLoading: isLoadingTypes,
+    error: typesError,
+  } = useQuery<AugmentationTypesResponse>({
+    queryKey: ['augmentation', 'types'],
+    queryFn: () => augmentationApi.getTypes(),
+    staleTime: 5 * 60 * 1000, // Cache for 5 minutes
+  })
+
+  // Fetch presets
+  const {
+    data: presetsData,
+    isLoading: isLoadingPresets,
+    error: presetsError,
+  } = useQuery<PresetsResponse>({
+    queryKey: ['augmentation', 'presets'],
+    queryFn: () => augmentationApi.getPresets(),
+    staleTime: 5 * 60 * 1000,
+  })
+
+  // Preview single augmentation mutation
+  const previewMutation = useMutation<PreviewResponse, Error, PreviewParams>({
+    mutationFn: ({ documentId, augmentationType, params, page = 1 }) =>
+      augmentationApi.preview(
+        documentId,
+        { augmentation_type: augmentationType, params },
+        page
+      ),
+    onError: (error) => {
+      console.error('Preview augmentation failed:', error)
+    },
+  })
+
+  // Preview full config mutation
+  const previewConfigMutation = useMutation<PreviewResponse, Error, PreviewConfigParams>({
+    mutationFn: ({ documentId, config, page = 1 }) =>
+      augmentationApi.previewConfig(documentId, config, page),
+    onError: (error) => {
+      console.error('Preview config failed:', error)
+    },
+  })
+
+  // Create augmented dataset mutation
+  const createBatchMutation = useMutation<BatchResponse, Error, BatchRequest>({
+    mutationFn: (request) => augmentationApi.createBatch(request),
+    onError: (error) => {
+      console.error('Create augmented dataset failed:', error)
+    },
+  })
+
+  return {
+    // Types data
+    augmentationTypes: typesData?.augmentation_types || [],
+    isLoadingTypes,
+    typesError,
+
+    // Presets data
+    presets: presetsData?.presets || [],
+    isLoadingPresets,
+    presetsError,
+
+    // Preview single augmentation
+    preview: previewMutation.mutate,
+    previewAsync: previewMutation.mutateAsync,
+    isPreviewing: previewMutation.isPending,
+    previewData: previewMutation.data,
+    previewError: previewMutation.error,
+
+    // Preview full config
+    previewConfig: previewConfigMutation.mutate,
+    previewConfigAsync: previewConfigMutation.mutateAsync,
+    isPreviewingConfig: previewConfigMutation.isPending,
+    previewConfigData: previewConfigMutation.data,
+    previewConfigError: previewConfigMutation.error,
+
+    // Create batch
+    createBatch: createBatchMutation.mutate,
+    createBatchAsync: createBatchMutation.mutateAsync,
+    isCreatingBatch: createBatchMutation.isPending,
+    batchData: createBatchMutation.data,
+    batchError: createBatchMutation.error,
+
+    // Reset functions for clearing stale mutation state
+    resetPreview: previewMutation.reset,
+    resetPreviewConfig: previewConfigMutation.reset,
+    resetBatch: createBatchMutation.reset,
+  }
+}
diff --git a/frontend/src/hooks/useDatasets.ts b/frontend/src/hooks/useDatasets.ts
new file mode 100644
index 0000000..a6fe6e0
--- /dev/null
+++ b/frontend/src/hooks/useDatasets.ts
@@ -0,0 +1,84 @@
+import { useQuery, useMutation, useQueryClient } from '@tanstack/react-query'
+import { datasetsApi } from '../api/endpoints'
+import type {
+  DatasetCreateRequest,
+  DatasetDetailResponse,
+  DatasetListResponse,
+  DatasetTrainRequest,
+} from '../api/types'
+
+export const useDatasets = (params?: {
+  status?: string
+  limit?: number
+  offset?: number
+}) => {
+  const queryClient = useQueryClient()
+
+  const { data, isLoading, error, refetch } = useQuery<DatasetListResponse>({
+    queryKey: ['datasets', params],
+    queryFn: () => datasetsApi.list(params),
+    staleTime: 30000,
+    // Poll every 5 seconds when there's an active training task
+    refetchInterval: (query) => {
+      const datasets = query.state.data?.datasets ?? []
+      const hasActiveTraining = datasets.some(
+        d => d.training_status === 'running' || d.training_status === 'pending' || d.training_status === 'scheduled'
+      )
+      return hasActiveTraining ? 5000 : false
+    },
+  })
+
+  const createMutation = useMutation({
+    mutationFn: (req: DatasetCreateRequest) => datasetsApi.create(req),
+    onSuccess: () => {
+      queryClient.invalidateQueries({ queryKey: ['datasets'] })
+    },
+  })
+
+  const deleteMutation = useMutation({
+    mutationFn: (datasetId: string) => datasetsApi.remove(datasetId),
+    onSuccess: () => {
+      queryClient.invalidateQueries({ queryKey: ['datasets'] })
+    },
+  })
+
+  const trainMutation = useMutation({
+    mutationFn: ({ datasetId, req }: { datasetId: string; req: DatasetTrainRequest }) =>
+      datasetsApi.trainFromDataset(datasetId, req),
+    onSuccess: () => {
+      queryClient.invalidateQueries({ queryKey: ['datasets'] })
+      queryClient.invalidateQueries({ queryKey: ['training', 'models'] })
+    },
+  })
+
+  return {
+    datasets: data?.datasets ?? [],
+    total: data?.total ?? 0,
+    isLoading,
+    error,
+    refetch,
+    createDataset: createMutation.mutate,
+    createDatasetAsync: createMutation.mutateAsync,
+    isCreating: createMutation.isPending,
+    deleteDataset: deleteMutation.mutate,
+    isDeleting: deleteMutation.isPending,
+    trainFromDataset: trainMutation.mutate,
+    trainFromDatasetAsync: trainMutation.mutateAsync,
+    isTraining: trainMutation.isPending,
+  }
+}
+
+export const useDatasetDetail = (datasetId: string | null) => {
+  const { data, isLoading, error } = useQuery<DatasetDetailResponse>({
+    queryKey: ['datasets', datasetId],
+    queryFn: () => datasetsApi.getDetail(datasetId!),
+    enabled: !!datasetId,
+    staleTime: 30000,
+  })
+
+  return {
+    dataset: data ?? null,
+    isLoading,
+    error,
+  }
+}
diff --git a/frontend/src/hooks/useDocuments.ts b/frontend/src/hooks/useDocuments.ts
index 22e07c1..b75a126 100644
--- a/frontend/src/hooks/useDocuments.ts
+++ b/frontend/src/hooks/useDocuments.ts
@@ -18,7 +18,16 @@ export const useDocuments = (params: UseDocumentsParams = {}) => {
   })
 
   const uploadMutation = useMutation({
-    mutationFn: (file: File) => documentsApi.upload(file),
+    mutationFn: ({ file, groupKey }: { file: File; groupKey?: string }) =>
+      documentsApi.upload(file, groupKey),
+    onSuccess: () => {
+      queryClient.invalidateQueries({ queryKey: ['documents'] })
+    },
+  })
+
+  const updateGroupKeyMutation = useMutation({
+    mutationFn: ({ documentId, groupKey }: { documentId: string; groupKey: string | null }) =>
+      documentsApi.updateGroupKey(documentId, groupKey),
     onSuccess: () => {
       queryClient.invalidateQueries({ queryKey: ['documents'] })
     },
@@ -74,5 +83,8 @@ export const useDocuments = (params: UseDocumentsParams = {}) => {
     isUpdatingStatus: updateStatusMutation.isPending,
     triggerAutoLabel: triggerAutoLabelMutation.mutate,
     isTriggeringAutoLabel: triggerAutoLabelMutation.isPending,
+    updateGroupKey: updateGroupKeyMutation.mutate,
+    updateGroupKeyAsync: updateGroupKeyMutation.mutateAsync,
+    isUpdatingGroupKey: updateGroupKeyMutation.isPending,
   }
 }
diff --git a/frontend/src/hooks/useModels.ts b/frontend/src/hooks/useModels.ts
new file mode 100644
index 0000000..9b3d8f7
--- /dev/null
+++ b/frontend/src/hooks/useModels.ts
@@ -0,0 +1,98 @@
+import { useQuery, useMutation, useQueryClient } from '@tanstack/react-query'
+import { modelsApi } from '../api/endpoints'
+import type {
+  ModelVersionListResponse,
+  ModelVersionDetailResponse,
+  ActiveModelResponse,
+} from '../api/types'
+
+export const useModels = (params?: {
+  status?: string
+  limit?: number
+  offset?: number
+}) => {
+  const queryClient = useQueryClient()
+
+  const { data, isLoading, error, refetch } = useQuery<ModelVersionListResponse>({
+    queryKey: ['models', params],
+    queryFn: () => modelsApi.list(params),
+    staleTime: 30000,
+  })
+
+  const activateMutation = useMutation({
+    mutationFn: (versionId: string) => modelsApi.activate(versionId),
+    onSuccess: () => {
+      queryClient.invalidateQueries({ queryKey: ['models'] })
+      queryClient.invalidateQueries({ queryKey: ['models', 'active'] })
+    },
+  })
+
+  const deactivateMutation = useMutation({
+    mutationFn: (versionId: string) => modelsApi.deactivate(versionId),
+    onSuccess: () => {
+      queryClient.invalidateQueries({ queryKey: ['models'] })
+      queryClient.invalidateQueries({ queryKey: ['models', 'active'] })
+    },
+  })
+
+  const archiveMutation = useMutation({
+    mutationFn: (versionId: string) => modelsApi.archive(versionId),
+    onSuccess: () => {
+      queryClient.invalidateQueries({ queryKey: ['models'] })
+    },
+  })
+
+  const deleteMutation = useMutation({
+    mutationFn: (versionId: string) => modelsApi.delete(versionId),
+    onSuccess: () => {
+      queryClient.invalidateQueries({ queryKey: ['models'] })
+    },
+  })
+
+  return {
+    models: data?.models ?? [],
+    total: data?.total ?? 0,
+    isLoading,
+    error,
+    refetch,
+    activateModel: activateMutation.mutate,
+    activateModelAsync: activateMutation.mutateAsync,
+    isActivating: activateMutation.isPending,
+    deactivateModel: deactivateMutation.mutate,
+    isDeactivating: deactivateMutation.isPending,
+    archiveModel: archiveMutation.mutate,
+    isArchiving: archiveMutation.isPending,
+    deleteModel: deleteMutation.mutate,
+    isDeleting: deleteMutation.isPending,
+  }
+}
+
+export const useModelDetail = (versionId: string | null) => {
+  const { data, isLoading, error } = useQuery<ModelVersionDetailResponse>({
+    queryKey: ['models', versionId],
+    queryFn: () => modelsApi.getDetail(versionId!),
+    enabled: !!versionId,
+    staleTime: 30000,
+  })
+
+  return {
+    model: data ?? null,
+    isLoading,
+    error,
+  }
+}
+
+export const useActiveModel = () => {
+  const { data, isLoading, error } = useQuery<ActiveModelResponse>({
+    queryKey: ['models', 'active'],
+    queryFn: () => modelsApi.getActive(),
+    staleTime: 30000,
+  })
+
+  return {
+    hasActiveModel: data?.has_active_model ?? false,
+    activeModel: data?.model ?? null,
+    isLoading,
+    error,
+  }
+}
diff --git a/migrations/005_add_group_key.sql b/migrations/005_add_group_key.sql
new file mode 100644
index 0000000..981caa4
--- /dev/null
+++ b/migrations/005_add_group_key.sql
@@ -0,0 +1,8 @@
+-- Add group_key column to admin_documents
+-- Allows users to organize documents into logical groups
+
+-- Add the column (nullable, VARCHAR 255)
+ALTER TABLE admin_documents ADD COLUMN IF NOT EXISTS group_key VARCHAR(255);
+
+-- Add index for filtering/grouping queries
+CREATE INDEX IF NOT EXISTS ix_admin_documents_group_key ON admin_documents(group_key);
diff --git a/migrations/006_model_versions.sql b/migrations/006_model_versions.sql
new file mode 100644
index 0000000..4d4e7f3
--- /dev/null
+++ b/migrations/006_model_versions.sql
@@ -0,0 +1,49 @@
+-- Model versions table for tracking trained model deployments.
+-- Each training run can produce a model version for inference.
+
+CREATE TABLE IF NOT EXISTS model_versions (
+    version_id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+    version VARCHAR(50) NOT NULL,
+    name VARCHAR(255) NOT NULL,
+    description TEXT,
+    model_path VARCHAR(512) NOT NULL,
+    status VARCHAR(20) NOT NULL DEFAULT 'inactive',
+    is_active BOOLEAN NOT NULL DEFAULT FALSE,
+
+    -- Training association
+    task_id UUID REFERENCES training_tasks(task_id) ON DELETE SET NULL,
+    dataset_id UUID REFERENCES training_datasets(dataset_id) ON DELETE SET NULL,
+
+    -- Training metrics
+    metrics_mAP DOUBLE PRECISION,
+    metrics_precision DOUBLE PRECISION,
+    metrics_recall DOUBLE PRECISION,
+    document_count INTEGER NOT NULL DEFAULT 0,
+
+    -- Training configuration snapshot
+    training_config JSONB,
+
+    -- File info
+    file_size BIGINT,
+
+    -- Timestamps
+    trained_at TIMESTAMP WITH TIME ZONE,
+    activated_at TIMESTAMP WITH TIME ZONE,
+    created_at TIMESTAMP WITH TIME ZONE NOT NULL DEFAULT NOW(),
+    updated_at TIMESTAMP WITH TIME ZONE NOT NULL DEFAULT NOW()
+);
+
+-- Indexes
+CREATE INDEX IF NOT EXISTS idx_model_versions_version ON model_versions(version);
+CREATE INDEX IF NOT EXISTS idx_model_versions_status ON model_versions(status);
+CREATE INDEX IF NOT EXISTS idx_model_versions_is_active ON model_versions(is_active);
+CREATE INDEX IF NOT EXISTS idx_model_versions_task_id ON model_versions(task_id);
+CREATE INDEX IF NOT EXISTS idx_model_versions_dataset_id ON model_versions(dataset_id);
+CREATE INDEX IF NOT EXISTS idx_model_versions_created ON model_versions(created_at);
+
+-- Ensure only one active model at a time
+CREATE UNIQUE INDEX IF NOT EXISTS idx_model_versions_single_active
+    ON model_versions(is_active) WHERE is_active = TRUE;
+
+-- Comment
+COMMENT ON TABLE model_versions IS 'Trained model versions for inference deployment';
diff --git a/migrations/007_training_tasks_extra_columns.sql b/migrations/007_training_tasks_extra_columns.sql
new file mode 100644
index 0000000..b4ef0ae
--- /dev/null
+++ b/migrations/007_training_tasks_extra_columns.sql
@@ -0,0 +1,46 @@
+-- Add missing columns to training_tasks table
+
+-- Add name column
+ALTER TABLE training_tasks ADD COLUMN IF NOT EXISTS name VARCHAR(255);
+UPDATE training_tasks SET name = 'Training ' || substring(task_id::text, 1, 8) WHERE name IS NULL;
+ALTER TABLE training_tasks ALTER COLUMN name SET NOT NULL;
+
+-- Add description column
+ALTER TABLE training_tasks ADD COLUMN IF NOT EXISTS description TEXT;
+
+-- Add admin_token column (for multi-tenant support)
+ALTER TABLE training_tasks ADD COLUMN IF NOT EXISTS admin_token VARCHAR(255);
+
+-- Add task_type column
+ALTER TABLE training_tasks ADD COLUMN IF NOT EXISTS task_type VARCHAR(20) DEFAULT 'train';
+
+-- Add recurring schedule columns
+ALTER TABLE training_tasks ADD COLUMN IF NOT EXISTS cron_expression VARCHAR(50);
+ALTER TABLE training_tasks ADD COLUMN IF NOT EXISTS is_recurring BOOLEAN DEFAULT FALSE;
+
+-- Add result metrics columns (for display without parsing JSONB)
+ALTER TABLE training_tasks ADD COLUMN IF NOT EXISTS result_metrics JSONB;
+ALTER TABLE training_tasks ADD COLUMN IF NOT EXISTS document_count INTEGER DEFAULT 0;
+ALTER TABLE training_tasks ADD COLUMN IF NOT EXISTS metrics_mAP DOUBLE PRECISION;
+ALTER TABLE training_tasks ADD COLUMN IF NOT EXISTS metrics_precision DOUBLE PRECISION;
+ALTER TABLE training_tasks ADD COLUMN IF NOT EXISTS metrics_recall DOUBLE PRECISION;
+
+-- Rename metrics to config if exists
+DO $$
+BEGIN
+    IF EXISTS (SELECT 1 FROM information_schema.columns
+               WHERE table_name = 'training_tasks' AND column_name = 'metrics'
+               AND NOT EXISTS (SELECT 1 FROM information_schema.columns
+                              WHERE table_name = 'training_tasks' AND column_name = 'config')) THEN
+        ALTER TABLE training_tasks RENAME COLUMN metrics TO config;
+    END IF;
+END $$;
+
+-- Add updated_at column
+ALTER TABLE training_tasks ADD COLUMN IF NOT EXISTS updated_at TIMESTAMP WITH TIME ZONE DEFAULT NOW();
+
+-- Create index on name
+CREATE INDEX IF NOT EXISTS idx_training_tasks_name ON training_tasks(name);
+
+-- Create index on metrics_mAP
+CREATE INDEX IF NOT EXISTS idx_training_tasks_mAP ON training_tasks(metrics_mAP);
diff --git a/migrations/008_fix_model_versions_fk.sql b/migrations/008_fix_model_versions_fk.sql
new file mode 100644
index 0000000..7a4dfc8
--- /dev/null
+++ b/migrations/008_fix_model_versions_fk.sql
@@ -0,0 +1,14 @@
+-- Fix foreign key constraints on model_versions table to allow CASCADE delete
+
+-- Drop existing constraints
+ALTER TABLE model_versions DROP CONSTRAINT IF EXISTS model_versions_dataset_id_fkey;
+ALTER TABLE model_versions DROP CONSTRAINT IF EXISTS model_versions_task_id_fkey;
+
+-- Add constraints with ON DELETE SET NULL
+ALTER TABLE model_versions
+ADD CONSTRAINT model_versions_dataset_id_fkey
+FOREIGN KEY (dataset_id) REFERENCES training_datasets(dataset_id) ON DELETE SET NULL;
+
+ALTER TABLE model_versions
+ADD CONSTRAINT model_versions_task_id_fkey
+FOREIGN KEY (task_id) REFERENCES training_tasks(task_id) ON DELETE SET NULL;
diff --git a/packages/inference/inference/data/admin_db.py b/packages/inference/inference/data/admin_db.py
index 02f9d8c..9f9765d 100644
--- a/packages/inference/inference/data/admin_db.py
+++ b/packages/inference/inference/data/admin_db.py
@@ -25,6 +25,7 @@ from inference.data.admin_models import (
     AnnotationHistory,
     TrainingDataset,
     DatasetDocument,
+    ModelVersion,
 )
 
 logger = logging.getLogger(__name__)
@@ -110,6 +111,7 @@ class AdminDB:
         page_count: int = 1,
         upload_source: str = "ui",
         csv_field_values: dict[str, Any] | None = None,
+        group_key: str | None = None,
         admin_token: str | None = None,  # Deprecated, kept for compatibility
     ) -> str:
         """Create a new document record."""
@@ -122,6 +124,7 @@ class AdminDB:
                 page_count=page_count,
                 upload_source=upload_source,
                 csv_field_values=csv_field_values,
+                group_key=group_key,
             )
             session.add(document)
             session.flush()
@@ -253,6 +256,17 @@ class AdminDB:
                 document.updated_at = datetime.utcnow()
                 session.add(document)
 
+    def update_document_group_key(self, document_id: str, group_key: str | None) -> bool:
+        """Update document group key."""
+        with get_session_context() as session:
+            document = session.get(AdminDocument, UUID(document_id))
+            if document:
+                document.group_key = group_key
+                document.updated_at = datetime.utcnow()
+                session.add(document)
+                return True
+            return False
+
     def delete_document(self, document_id: str) -> bool:
         """Delete a document and its annotations."""
         with get_session_context() as session:
@@ -1215,6 +1229,39 @@ class AdminDB:
                 session.expunge(d)
             return list(datasets), total
 
+    def get_active_training_tasks_for_datasets(
+        self, dataset_ids: list[str]
+    ) -> dict[str, dict[str, str]]:
+        """Get active (pending/scheduled/running) training tasks for datasets.
+
+        Returns a dict mapping dataset_id to {"task_id": ..., "status": ...}
+        """
+        if not dataset_ids:
+            return {}
+
+        # Validate UUIDs before query
+        valid_uuids = []
+        for d in dataset_ids:
+            try:
+                valid_uuids.append(UUID(d))
+            except ValueError:
+                logger.warning("Invalid UUID in get_active_training_tasks_for_datasets: %s", d)
+                continue
+
+        if not valid_uuids:
+            return {}
+
+        with get_session_context() as session:
+            statement = select(TrainingTask).where(
+                TrainingTask.dataset_id.in_(valid_uuids),
+                TrainingTask.status.in_(["pending", "scheduled", "running"]),
+            )
+            results = session.exec(statement).all()
+            return {
+                str(t.dataset_id): {"task_id": str(t.task_id), "status": t.status}
+                for t in results
+            }
+
     def update_dataset_status(
         self,
         dataset_id: str | UUID,
@@ -1314,3 +1361,182 @@ class AdminDB:
             session.delete(dataset)
             session.commit()
             return True
+
+    # ==========================================================================
+    # Model Version Operations
+    # ==========================================================================
+
+    def create_model_version(
+        self,
+        version: str,
+        name: str,
+        model_path: str,
+        description: str | None = None,
+        task_id: str | UUID | None = None,
+        dataset_id: str | UUID | None = None,
+        metrics_mAP: float | None = None,
+        metrics_precision: float | None = None,
+        metrics_recall: float | None = None,
+        document_count: int = 0,
+        training_config: dict[str, Any] | None = None,
+        file_size: int | None = None,
+        trained_at: datetime | None = None,
+    ) -> ModelVersion:
+        """Create a new model version."""
+        with get_session_context() as session:
+            model = ModelVersion(
+                version=version,
+                name=name,
+                model_path=model_path,
+                description=description,
+                task_id=UUID(str(task_id)) if task_id else None,
+                dataset_id=UUID(str(dataset_id)) if dataset_id else None,
+                metrics_mAP=metrics_mAP,
+                metrics_precision=metrics_precision,
+                metrics_recall=metrics_recall,
+                document_count=document_count,
+                training_config=training_config,
+                file_size=file_size,
+                trained_at=trained_at,
+            )
+            session.add(model)
+            session.commit()
+            session.refresh(model)
+            session.expunge(model)
+            return model
+
+    def get_model_version(self, version_id: str | UUID) -> ModelVersion | None:
+        """Get a model version by ID."""
+        with get_session_context() as session:
+            model = session.get(ModelVersion, UUID(str(version_id)))
+            if model:
+                session.expunge(model)
+            return model
+
+    def get_model_versions(
+        self,
+        status: str | None = None,
+        limit: int = 20,
+        offset: int = 0,
+    ) -> tuple[list[ModelVersion], int]:
+        """List model versions with optional status filter."""
+        with get_session_context() as session:
+            query = select(ModelVersion)
+            count_query = select(func.count()).select_from(ModelVersion)
+            if status:
+                query = query.where(ModelVersion.status == status)
+                count_query = count_query.where(ModelVersion.status == status)
+            total = session.exec(count_query).one()
+            models = session.exec(
+                query.order_by(ModelVersion.created_at.desc()).offset(offset).limit(limit)
+            ).all()
+            for m in models:
+                session.expunge(m)
+            return list(models), total
+
+    def get_active_model_version(self) -> ModelVersion | None:
+        """Get the currently active model version for inference."""
+        with get_session_context() as session:
+            result = session.exec(
+                select(ModelVersion).where(ModelVersion.is_active == True)
+            ).first()
+            if result:
+                session.expunge(result)
+            return result
+
+    def activate_model_version(self, version_id: str | UUID) -> ModelVersion | None:
+        """Activate a model version for inference (deactivates all others)."""
+        with get_session_context() as session:
+            # Deactivate all versions
+            all_versions = session.exec(
+                select(ModelVersion).where(ModelVersion.is_active == True)
+            ).all()
+            for v in all_versions:
+                v.is_active = False
+                v.status = "inactive"
+                v.updated_at = datetime.utcnow()
+                session.add(v)
+
+            # Activate the specified version
+            model = session.get(ModelVersion, UUID(str(version_id)))
+            if not model:
+                return None
+            model.is_active = True
+            model.status = "active"
+            model.activated_at = datetime.utcnow()
+            model.updated_at = datetime.utcnow()
+            session.add(model)
+            session.commit()
+            session.refresh(model)
+            session.expunge(model)
+            return model
+
+    def deactivate_model_version(self, version_id: str | UUID) -> ModelVersion | None:
+        """Deactivate a model version."""
+        with get_session_context() as session:
+            model = session.get(ModelVersion, UUID(str(version_id)))
+            if not model:
+                return None
+            model.is_active = False
+            model.status = "inactive"
+            model.updated_at = datetime.utcnow()
+            session.add(model)
+            session.commit()
+            session.refresh(model)
+            session.expunge(model)
+            return model
+
+    def update_model_version(
+        self,
+        version_id: str | UUID,
+        name: str | None = None,
+        description: str | None = None,
+        status: str | None = None,
+    ) -> ModelVersion | None:
+        """Update model version metadata."""
+        with get_session_context() as session:
+            model = session.get(ModelVersion, UUID(str(version_id)))
+            if not model:
+                return None
+            if name is not None:
+                model.name = name
+            if description is not None:
+                model.description = description
+            if status is not None:
+                model.status = status
+            model.updated_at = datetime.utcnow()
+            session.add(model)
+            session.commit()
+            session.refresh(model)
+            session.expunge(model)
+            return model
+
+    def archive_model_version(self, version_id: str | UUID) -> ModelVersion | None:
+        """Archive a model version."""
+        with get_session_context() as session:
+            model = session.get(ModelVersion, UUID(str(version_id)))
+            if not model:
+                return None
+            # Cannot archive active model
+            if model.is_active:
+                return None
+            model.status = "archived"
+            model.updated_at = datetime.utcnow()
+            session.add(model)
+            session.commit()
+            session.refresh(model)
+            session.expunge(model)
+            return model
+
+    def delete_model_version(self, version_id: str | UUID) -> bool:
+        """Delete a model version."""
+        with get_session_context() as session:
+            model = session.get(ModelVersion, UUID(str(version_id)))
+            if not model:
+                return False
+            # Cannot delete active model
+            if model.is_active:
+                return False
+            session.delete(model)
+            session.commit()
+            return True
diff --git a/packages/inference/inference/data/admin_models.py b/packages/inference/inference/data/admin_models.py
index a374d07..ca7e5f2 100644
--- a/packages/inference/inference/data/admin_models.py
+++ b/packages/inference/inference/data/admin_models.py
@@ -70,6 +70,8 @@ class AdminDocument(SQLModel, table=True):
     # Upload source: ui, api
     batch_id: UUID | None = Field(default=None, index=True)
     # Link to batch upload (if uploaded via ZIP)
+    group_key: str | None = Field(default=None, max_length=255, index=True)
+    # User-defined grouping key for document organization
     csv_field_values: dict[str, Any] | None = Field(default=None, sa_column=Column(JSON))
     # Original CSV values for reference
     auto_label_queued_at: datetime | None = Field(default=None)
@@ -275,6 +277,56 @@ class TrainingDocumentLink(SQLModel, table=True):
     created_at: datetime = Field(default_factory=datetime.utcnow)
 
 
+# =============================================================================
+# Model Version Management
+# =============================================================================
+
+
+class ModelVersion(SQLModel, table=True):
+    """Model version for inference deployment."""
+
+    __tablename__ = "model_versions"
+
+    version_id: UUID = Field(default_factory=uuid4, primary_key=True)
+    version: str = Field(max_length=50, index=True)
+    # Semantic version e.g., "1.0.0", "2.1.0"
+    name: str = Field(max_length=255)
+    description: str | None = Field(default=None)
+    model_path: str = Field(max_length=512)
+    # Path to the model weights file
+    status: str = Field(default="inactive", max_length=20, index=True)
+    # Status: active, inactive, archived
+    is_active: bool = Field(default=False, index=True)
+    # Only one version can be active at a time for inference
+
+    # Training association
+    task_id: UUID | None = Field(default=None, foreign_key="training_tasks.task_id", index=True)
+    dataset_id: UUID | None = Field(default=None, foreign_key="training_datasets.dataset_id", index=True)
+
+    # Training metrics
+    metrics_mAP: float | None = Field(default=None)
+    metrics_precision: float | None = Field(default=None)
+    metrics_recall: float | None = Field(default=None)
+    document_count: int = Field(default=0)
+    # Number of documents used in training
+
+    # Training configuration snapshot
+    training_config: dict[str, Any] | None = Field(default=None, sa_column=Column(JSON))
+    # Snapshot of epochs, batch_size, etc.
+
+    # File info
+    file_size: int | None = Field(default=None)
+    # Model file size in bytes
+
+    # Timestamps
+    trained_at: datetime | None = Field(default=None)
+    # When training completed
+    activated_at: datetime | None = Field(default=None)
+    # When this version was last activated
+    created_at: datetime = Field(default_factory=datetime.utcnow)
+    updated_at: datetime = Field(default_factory=datetime.utcnow)
+
+
 # =============================================================================
 # Annotation History (v2)
 # =============================================================================
diff --git a/packages/inference/inference/data/database.py b/packages/inference/inference/data/database.py
index 7613b6f..15b4c14 100644
--- a/packages/inference/inference/data/database.py
+++ b/packages/inference/inference/data/database.py
@@ -49,6 +49,111 @@ def get_engine():
     return _engine
 
 
+def run_migrations() -> None:
+    """Run database migrations for new columns."""
+    engine = get_engine()
+
+    migrations = [
+        # Migration 004: Training datasets tables and dataset_id on training_tasks
+        (
+            "training_datasets_tables",
+            """
+            CREATE TABLE IF NOT EXISTS training_datasets (
+                dataset_id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+                name VARCHAR(255) NOT NULL,
+                description TEXT,
+                status VARCHAR(20) NOT NULL DEFAULT 'building',
+                train_ratio FLOAT NOT NULL DEFAULT 0.8,
+                val_ratio FLOAT NOT NULL DEFAULT 0.1,
+                seed INTEGER NOT NULL DEFAULT 42,
+                total_documents INTEGER NOT NULL DEFAULT 0,
+                total_images INTEGER NOT NULL DEFAULT 0,
+                total_annotations INTEGER NOT NULL DEFAULT 0,
+                dataset_path VARCHAR(512),
+                error_message TEXT,
+                created_at TIMESTAMP WITH TIME ZONE NOT NULL DEFAULT NOW(),
+                updated_at TIMESTAMP WITH TIME ZONE NOT NULL DEFAULT NOW()
+            );
+            CREATE INDEX IF NOT EXISTS idx_training_datasets_status ON training_datasets(status);
+            """,
+        ),
+        (
+            "dataset_documents_table",
+            """
+            CREATE TABLE IF NOT EXISTS dataset_documents (
+                id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+                dataset_id UUID NOT NULL REFERENCES training_datasets(dataset_id) ON DELETE CASCADE,
+                document_id UUID NOT NULL REFERENCES admin_documents(document_id),
+                split VARCHAR(10) NOT NULL,
+                page_count INTEGER NOT NULL DEFAULT 0,
+                annotation_count INTEGER NOT NULL DEFAULT 0,
+                created_at TIMESTAMP WITH TIME ZONE NOT NULL DEFAULT NOW(),
+                UNIQUE(dataset_id, document_id)
+            );
+            CREATE INDEX IF NOT EXISTS idx_dataset_documents_dataset ON dataset_documents(dataset_id);
+            CREATE INDEX IF NOT EXISTS idx_dataset_documents_document ON dataset_documents(document_id);
+            """,
+        ),
+        (
+            "training_tasks_dataset_id",
+            """
+            ALTER TABLE training_tasks ADD COLUMN IF NOT EXISTS dataset_id UUID REFERENCES training_datasets(dataset_id);
+            CREATE INDEX IF NOT EXISTS idx_training_tasks_dataset ON training_tasks(dataset_id);
+            """,
+        ),
+        # Migration 005: Add group_key to admin_documents
+        (
+            "admin_documents_group_key",
+            """
+            ALTER TABLE admin_documents ADD COLUMN IF NOT EXISTS group_key VARCHAR(255);
+            CREATE INDEX IF NOT EXISTS ix_admin_documents_group_key ON admin_documents(group_key);
+            """,
+        ),
+        # Migration 006: Model versions table
+        (
+            "model_versions_table",
+            """
+            CREATE TABLE IF NOT EXISTS model_versions (
+                version_id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+                version VARCHAR(50) NOT NULL,
+                name VARCHAR(255) NOT NULL,
+                description TEXT,
+                model_path VARCHAR(512) NOT NULL,
+                status VARCHAR(20) NOT NULL DEFAULT 'inactive',
+                is_active BOOLEAN NOT NULL DEFAULT FALSE,
+                task_id UUID REFERENCES training_tasks(task_id),
+                dataset_id UUID REFERENCES training_datasets(dataset_id),
+                metrics_mAP FLOAT,
+                metrics_precision FLOAT,
+                metrics_recall FLOAT,
+                document_count INTEGER NOT NULL DEFAULT 0,
+                training_config JSONB,
+                file_size BIGINT,
+                trained_at TIMESTAMP WITH TIME ZONE,
+                activated_at TIMESTAMP WITH TIME ZONE,
+                created_at TIMESTAMP WITH TIME ZONE NOT NULL DEFAULT NOW(),
+                updated_at TIMESTAMP WITH TIME ZONE NOT NULL DEFAULT NOW()
+            );
+            CREATE INDEX IF NOT EXISTS ix_model_versions_version ON model_versions(version);
+            CREATE INDEX IF NOT EXISTS ix_model_versions_status ON model_versions(status);
+            CREATE INDEX IF NOT EXISTS ix_model_versions_is_active ON model_versions(is_active);
+            CREATE INDEX IF NOT EXISTS ix_model_versions_task_id ON model_versions(task_id);
+            CREATE INDEX IF NOT EXISTS ix_model_versions_dataset_id ON model_versions(dataset_id);
+            """,
+        ),
+    ]
+
+    with engine.connect() as conn:
+        for name, sql in migrations:
+            try:
+                conn.execute(text(sql))
+                conn.commit()
+                logger.info(f"Migration '{name}' applied successfully")
+            except Exception as e:
+                # Log but don't fail - column may already exist
+                logger.debug(f"Migration '{name}' skipped or failed: {e}")
+
+
 def create_db_and_tables() -> None:
     """Create all database tables."""
     from inference.data.models import ApiKey, AsyncRequest, RateLimitEvent  # noqa: F401
@@ -64,6 +169,9 @@ def create_db_and_tables() -> None:
     SQLModel.metadata.create_all(engine)
     logger.info("Database tables created/verified")
 
+    # Run migrations for new columns
+    run_migrations()
+
 
 def get_session() -> Session:
     """Get a new database session."""
diff --git a/packages/inference/inference/web/api/v1/admin/__init__.py b/packages/inference/inference/web/api/v1/admin/__init__.py
index 8d5081b..ce7a5d0 100644
--- a/packages/inference/inference/web/api/v1/admin/__init__.py
+++ b/packages/inference/inference/web/api/v1/admin/__init__.py
@@ -5,6 +5,7 @@ Document management, annotations, and training endpoints.
 """
 
 from inference.web.api.v1.admin.annotations import create_annotation_router
+from inference.web.api.v1.admin.augmentation import create_augmentation_router
 from inference.web.api.v1.admin.auth import create_auth_router
 from inference.web.api.v1.admin.documents import create_documents_router
 from inference.web.api.v1.admin.locks import create_locks_router
@@ -12,6 +13,7 @@ from inference.web.api.v1.admin.training import create_training_router
 
 __all__ = [
     "create_annotation_router",
+    "create_augmentation_router",
     "create_auth_router",
     "create_documents_router",
     "create_locks_router",
diff --git a/packages/inference/inference/web/api/v1/admin/augmentation/__init__.py b/packages/inference/inference/web/api/v1/admin/augmentation/__init__.py
new file mode 100644
index 0000000..185e649
--- /dev/null
+++ b/packages/inference/inference/web/api/v1/admin/augmentation/__init__.py
@@ -0,0 +1,15 @@
+"""Augmentation API module."""
+
+from fastapi import APIRouter
+
+from .routes import register_augmentation_routes
+
+
+def create_augmentation_router() -> APIRouter:
+    """Create and configure the augmentation router."""
+    router = APIRouter(prefix="/augmentation", tags=["augmentation"])
+    register_augmentation_routes(router)
+    return router
+
+
+__all__ = ["create_augmentation_router"]
diff --git a/packages/inference/inference/web/api/v1/admin/augmentation/routes.py b/packages/inference/inference/web/api/v1/admin/augmentation/routes.py
new file mode 100644
index 0000000..fbf6e3e
--- /dev/null
+++ b/packages/inference/inference/web/api/v1/admin/augmentation/routes.py
@@ -0,0 +1,162 @@
+"""Augmentation API routes."""
+
+from typing import Annotated
+
+from fastapi import APIRouter, HTTPException, Query
+
+from inference.web.core.auth import AdminDBDep, AdminTokenDep
+from inference.web.schemas.admin.augmentation import (
+    AugmentationBatchRequest,
+    AugmentationBatchResponse,
+    AugmentationConfigSchema,
+    AugmentationPreviewRequest,
+    AugmentationPreviewResponse,
+    AugmentationTypeInfo,
+    AugmentationTypesResponse,
+    AugmentedDatasetItem,
+    AugmentedDatasetListResponse,
+    PresetInfo,
+    PresetsResponse,
+)
+
+
+def register_augmentation_routes(router: APIRouter) -> None:
+    """Register augmentation endpoints on the router."""
+
+    @router.get(
+        "/types",
+        response_model=AugmentationTypesResponse,
+        summary="List available augmentation types",
+    )
+    async def list_augmentation_types(
+        admin_token: AdminTokenDep,
+    ) -> AugmentationTypesResponse:
+        """
+        List all available augmentation types with descriptions and parameters.
+        """
+        from shared.augmentation.pipeline import (
+            AUGMENTATION_REGISTRY,
+            AugmentationPipeline,
+        )
+
+        types = []
+        for name, aug_class in AUGMENTATION_REGISTRY.items():
+            # Create instance with empty params to get preview params
+            aug = aug_class({})
+            types.append(
+                AugmentationTypeInfo(
+                    name=name,
+                    description=(aug_class.__doc__ or "").strip(),
+                    affects_geometry=aug_class.affects_geometry,
+                    stage=AugmentationPipeline.STAGE_MAPPING[name],
+                    default_params=aug.get_preview_params(),
+                )
+            )
+
+        return AugmentationTypesResponse(augmentation_types=types)
+
+    @router.get(
+        "/presets",
+        response_model=PresetsResponse,
+        summary="Get augmentation presets",
+    )
+    async def get_presets(
+        admin_token: AdminTokenDep,
+    ) -> PresetsResponse:
+        """Get predefined augmentation presets for common use cases."""
+        from shared.augmentation.presets import list_presets
+
+        presets = [PresetInfo(**p) for p in list_presets()]
+        return PresetsResponse(presets=presets)
+
+    @router.post(
+        "/preview/{document_id}",
+        response_model=AugmentationPreviewResponse,
+        summary="Preview augmentation on document image",
+    )
+    async def preview_augmentation(
+        document_id: str,
+        request: AugmentationPreviewRequest,
+        admin_token: AdminTokenDep,
+        db: AdminDBDep,
+        page: int = Query(default=1, ge=1, description="Page number"),
+    ) -> AugmentationPreviewResponse:
+        """
+        Preview a single augmentation on a document page.
+
+        Returns URLs to original and augmented preview images.
+        """
+        from inference.web.services.augmentation_service import AugmentationService
+
+        service = AugmentationService(db=db)
+        return await service.preview_single(
+            document_id=document_id,
+            page=page,
+            augmentation_type=request.augmentation_type,
+            params=request.params,
+        )
+
+    @router.post(
+        "/preview-config/{document_id}",
+        response_model=AugmentationPreviewResponse,
+        summary="Preview full augmentation config on document",
+    )
+    async def preview_config(
+        document_id: str,
+        config: AugmentationConfigSchema,
+        admin_token: AdminTokenDep,
+        db: AdminDBDep,
+        page: int = Query(default=1, ge=1, description="Page number"),
+    ) -> AugmentationPreviewResponse:
+        """Preview complete augmentation pipeline on a document page."""
+        from inference.web.services.augmentation_service import AugmentationService
+
+        service = AugmentationService(db=db)
+        return await service.preview_config(
+            document_id=document_id,
+            page=page,
+            config=config,
+        )
+
+    @router.post(
+        "/batch",
+        response_model=AugmentationBatchResponse,
+        summary="Create augmented dataset (offline preprocessing)",
+    )
+    async def create_augmented_dataset(
+        request: AugmentationBatchRequest,
+        admin_token: AdminTokenDep,
+        db: AdminDBDep,
+    ) -> AugmentationBatchResponse:
+        """
+        Create a new augmented dataset from an existing dataset.
+
+        This runs as a background task. The augmented images are stored
+        alongside the original dataset for training.
+        """
+        from inference.web.services.augmentation_service import AugmentationService
+
+        service = AugmentationService(db=db)
+        return await service.create_augmented_dataset(
+            source_dataset_id=request.dataset_id,
+            config=request.config,
+            output_name=request.output_name,
+            multiplier=request.multiplier,
+        )
+
+    @router.get(
+        "/datasets",
+        response_model=AugmentedDatasetListResponse,
+        summary="List augmented datasets",
+    )
+    async def list_augmented_datasets(
+        admin_token: AdminTokenDep,
+        db: AdminDBDep,
+        limit: int = Query(default=20, ge=1, le=100, description="Page size"),
+        offset: int = Query(default=0, ge=0, description="Offset"),
+    ) -> AugmentedDatasetListResponse:
+        """List all augmented datasets."""
+        from inference.web.services.augmentation_service import AugmentationService
+
+        service = AugmentationService(db=db)
+        return await service.list_augmented_datasets(limit=limit, offset=offset)
diff --git a/packages/inference/inference/web/api/v1/admin/documents.py b/packages/inference/inference/web/api/v1/admin/documents.py
index fd2f355..f78db66 100644
--- a/packages/inference/inference/web/api/v1/admin/documents.py
+++ b/packages/inference/inference/web/api/v1/admin/documents.py
@@ -91,8 +91,19 @@ def create_documents_router(storage_config: StorageConfig) -> APIRouter:
             bool,
             Query(description="Trigger auto-labeling after upload"),
         ] = True,
+        group_key: Annotated[
+            str | None,
+            Query(description="Optional group key for document organization", max_length=255),
+        ] = None,
     ) -> DocumentUploadResponse:
         """Upload a document for labeling."""
+        # Validate group_key length
+        if group_key and len(group_key) > 255:
+            raise HTTPException(
+                status_code=400,
+                detail="Group key must be 255 characters or less",
+            )
+
         # Validate filename
         if not file.filename:
             raise HTTPException(status_code=400, detail="Filename is required")
@@ -131,6 +142,7 @@ def create_documents_router(storage_config: StorageConfig) -> APIRouter:
             content_type=file.content_type or "application/octet-stream",
             file_path="",  # Will update after saving
             page_count=page_count,
+            group_key=group_key,
         )
 
         # Save file to admin uploads
@@ -177,6 +189,7 @@ def create_documents_router(storage_config: StorageConfig) -> APIRouter:
             file_size=len(content),
             page_count=page_count,
             status=DocumentStatus.AUTO_LABELING if auto_label_started else DocumentStatus.PENDING,
+            group_key=group_key,
             auto_label_started=auto_label_started,
             message="Document uploaded successfully",
         )
@@ -277,6 +290,7 @@ def create_documents_router(storage_config: StorageConfig) -> APIRouter:
                     annotation_count=len(annotations),
                     upload_source=doc.upload_source if hasattr(doc, 'upload_source') else "ui",
                     batch_id=str(doc.batch_id) if hasattr(doc, 'batch_id') and doc.batch_id else None,
+                    group_key=doc.group_key if hasattr(doc, 'group_key') else None,
                     can_annotate=can_annotate,
                     created_at=doc.created_at,
                     updated_at=doc.updated_at,
@@ -421,6 +435,7 @@ def create_documents_router(storage_config: StorageConfig) -> APIRouter:
             auto_label_error=document.auto_label_error,
             upload_source=document.upload_source if hasattr(document, 'upload_source') else "ui",
             batch_id=str(document.batch_id) if hasattr(document, 'batch_id') and document.batch_id else None,
+            group_key=document.group_key if hasattr(document, 'group_key') else None,
             csv_field_values=csv_field_values,
             can_annotate=can_annotate,
             annotation_lock_until=annotation_lock_until,
@@ -548,4 +563,50 @@ def create_documents_router(storage_config: StorageConfig) -> APIRouter:
 
         return response
 
+    @router.patch(
+        "/{document_id}/group-key",
+        responses={
+            401: {"model": ErrorResponse, "description": "Invalid token"},
+            404: {"model": ErrorResponse, "description": "Document not found"},
+        },
+        summary="Update document group key",
+        description="Update the group key for a document.",
+    )
+    async def update_document_group_key(
+        document_id: str,
+        admin_token: AdminTokenDep,
+        db: AdminDBDep,
+        group_key: Annotated[
+            str | None,
+            Query(description="New group key (null to clear)"),
+        ] = None,
+    ) -> dict:
+        """Update document group key."""
+        _validate_uuid(document_id, "document_id")
+
+        # Validate group_key length
+        if group_key and len(group_key) > 255:
+            raise HTTPException(
+                status_code=400,
+                detail="Group key must be 255 characters or less",
+            )
+
+        # Verify document exists
+        document = db.get_document_by_token(document_id, admin_token)
+        if document is None:
+            raise HTTPException(
+                status_code=404,
+                detail="Document not found or does not belong to this token",
+            )
+
+        # Update group key
+        db.update_document_group_key(document_id, group_key)
+
+        return {
+            "status": "updated",
+            "document_id": document_id,
+            "group_key": group_key,
+            "message": "Document group key updated",
+        }
+
     return router
diff --git a/packages/inference/inference/web/api/v1/admin/training/__init__.py b/packages/inference/inference/web/api/v1/admin/training/__init__.py
index d2fba3c..cde7547 100644
--- a/packages/inference/inference/web/api/v1/admin/training/__init__.py
+++ b/packages/inference/inference/web/api/v1/admin/training/__init__.py
@@ -11,6 +11,7 @@ from .tasks import register_task_routes
 from .documents import register_document_routes
 from .export import register_export_routes
 from .datasets import register_dataset_routes
+from .models import register_model_routes
 
 
 def create_training_router() -> APIRouter:
@@ -21,6 +22,7 @@ def create_training_router() -> APIRouter:
     register_document_routes(router)
     register_export_routes(router)
     register_dataset_routes(router)
+    register_model_routes(router)
 
     return router
 
diff --git a/packages/inference/inference/web/api/v1/admin/training/datasets.py b/packages/inference/inference/web/api/v1/admin/training/datasets.py
index a46c4b3..bf93239 100644
--- a/packages/inference/inference/web/api/v1/admin/training/datasets.py
+++ b/packages/inference/inference/web/api/v1/admin/training/datasets.py
@@ -41,6 +41,13 @@ def register_dataset_routes(router: APIRouter) -> None:
         from pathlib import Path
         from inference.web.services.dataset_builder import DatasetBuilder
 
+        # Validate minimum document count for proper train/val/test split
+        if len(request.document_ids) < 10:
+            raise HTTPException(
+                status_code=400,
+                detail=f"Minimum 10 documents required for training dataset (got {len(request.document_ids)})",
+            )
+
         dataset = db.create_dataset(
             name=request.name,
             description=request.description,
@@ -83,6 +90,15 @@ def register_dataset_routes(router: APIRouter) -> None:
     ) -> DatasetListResponse:
         """List training datasets."""
         datasets, total = db.get_datasets(status=status, limit=limit, offset=offset)
+
+        # Get active training tasks for each dataset (graceful degradation on error)
+        dataset_ids = [str(d.dataset_id) for d in datasets]
+        try:
+            active_tasks = db.get_active_training_tasks_for_datasets(dataset_ids)
+        except Exception:
+            logger.exception("Failed to fetch active training tasks")
+            active_tasks = {}
+
         return DatasetListResponse(
             total=total,
             limit=limit,
@@ -93,6 +109,8 @@ def register_dataset_routes(router: APIRouter) -> None:
                     name=d.name,
                     description=d.description,
                     status=d.status,
+                    training_status=active_tasks.get(str(d.dataset_id), {}).get("status"),
+                    active_training_task_id=active_tasks.get(str(d.dataset_id), {}).get("task_id"),
                     total_documents=d.total_documents,
                     total_images=d.total_images,
                     total_annotations=d.total_annotations,
@@ -175,6 +193,7 @@ def register_dataset_routes(router: APIRouter) -> None:
         "/datasets/{dataset_id}/train",
         response_model=TrainingTaskResponse,
         summary="Start training from dataset",
+        description="Create a training task. Set base_model_version_id in config for incremental training.",
     )
     async def train_from_dataset(
         dataset_id: str,
@@ -182,7 +201,11 @@ def register_dataset_routes(router: APIRouter) -> None:
         admin_token: AdminTokenDep,
         db: AdminDBDep,
     ) -> TrainingTaskResponse:
-        """Create a training task from a dataset."""
+        """Create a training task from a dataset.
+
+        For incremental training, set config.base_model_version_id to a model version UUID.
+        The training will use that model as the starting point instead of a pretrained model.
+        """
         _validate_uuid(dataset_id, "dataset_id")
         dataset = db.get_dataset(dataset_id)
         if not dataset:
@@ -194,16 +217,42 @@ def register_dataset_routes(router: APIRouter) -> None:
             )
 
         config_dict = request.config.model_dump()
+
+        # Resolve base_model_version_id to actual model path for incremental training
+        base_model_version_id = config_dict.get("base_model_version_id")
+        if base_model_version_id:
+            _validate_uuid(base_model_version_id, "base_model_version_id")
+            base_model = db.get_model_version(base_model_version_id)
+            if not base_model:
+                raise HTTPException(
+                    status_code=404,
+                    detail=f"Base model version not found: {base_model_version_id}",
+                )
+            # Store the resolved model path for the training worker
+            config_dict["base_model_path"] = base_model.model_path
+            config_dict["base_model_version"] = base_model.version
+            logger.info(
+                "Incremental training: using model %s (%s) as base",
+                base_model.version,
+                base_model.model_path,
+            )
+
         task_id = db.create_training_task(
             admin_token=admin_token,
             name=request.name,
-            task_type="train",
+            task_type="finetune" if base_model_version_id else "train",
             config=config_dict,
             dataset_id=str(dataset.dataset_id),
         )
 
+        message = (
+            f"Incremental training task created (base: v{config_dict.get('base_model_version', 'N/A')})"
+            if base_model_version_id
+            else "Training task created from dataset"
+        )
+
         return TrainingTaskResponse(
             task_id=task_id,
             status=TrainingStatus.PENDING,
-            message="Training task created from dataset",
+            message=message,
         )
diff --git a/packages/inference/inference/web/api/v1/admin/training/documents.py b/packages/inference/inference/web/api/v1/admin/training/documents.py
index 27e935a..18e9e7d 100644
--- a/packages/inference/inference/web/api/v1/admin/training/documents.py
+++ b/packages/inference/inference/web/api/v1/admin/training/documents.py
@@ -145,15 +145,15 @@ def register_document_routes(router: APIRouter) -> None:
         )
 
     @router.get(
-        "/models",
+        "/completed-tasks",
         response_model=TrainingModelsResponse,
         responses={
             401: {"model": ErrorResponse, "description": "Invalid token"},
         },
-        summary="Get trained models",
-        description="Get list of trained models with metrics and download links.",
+        summary="Get completed training tasks",
+        description="Get list of completed training tasks with metrics and download links. For model versions, use /models endpoint.",
     )
-    async def get_training_models(
+    async def get_completed_training_tasks(
         admin_token: AdminTokenDep,
         db: AdminDBDep,
         status: Annotated[
diff --git a/packages/inference/inference/web/api/v1/admin/training/models.py b/packages/inference/inference/web/api/v1/admin/training/models.py
new file mode 100644
index 0000000..fcbb64b
--- /dev/null
+++ b/packages/inference/inference/web/api/v1/admin/training/models.py
@@ -0,0 +1,333 @@
+"""Model Version Endpoints."""
+
+import logging
+from typing import Annotated
+
+from fastapi import APIRouter, HTTPException, Query, Request
+
+from inference.web.core.auth import AdminTokenDep, AdminDBDep
+from inference.web.schemas.admin import (
+    ModelVersionCreateRequest,
+    ModelVersionUpdateRequest,
+    ModelVersionItem,
+    ModelVersionListResponse,
+    ModelVersionDetailResponse,
+    ModelVersionResponse,
+    ActiveModelResponse,
+)
+
+from ._utils import _validate_uuid
+
+logger = logging.getLogger(__name__)
+
+
+def register_model_routes(router: APIRouter) -> None:
+    """Register model version endpoints on the router."""
+
+    @router.post(
+        "/models",
+        response_model=ModelVersionResponse,
+        summary="Create model version",
+        description="Register a new model version for deployment.",
+    )
+    async def create_model_version(
+        request: ModelVersionCreateRequest,
+        admin_token: AdminTokenDep,
+        db: AdminDBDep,
+    ) -> ModelVersionResponse:
+        """Create a new model version."""
+        if request.task_id:
+            _validate_uuid(request.task_id, "task_id")
+        if request.dataset_id:
+            _validate_uuid(request.dataset_id, "dataset_id")
+
+        model = db.create_model_version(
+            version=request.version,
+            name=request.name,
+            model_path=request.model_path,
+            description=request.description,
+            task_id=request.task_id,
+            dataset_id=request.dataset_id,
+            metrics_mAP=request.metrics_mAP,
+            metrics_precision=request.metrics_precision,
+            metrics_recall=request.metrics_recall,
+            document_count=request.document_count,
+            training_config=request.training_config,
+            file_size=request.file_size,
+            trained_at=request.trained_at,
+        )
+
+        return ModelVersionResponse(
+            version_id=str(model.version_id),
+            status=model.status,
+            message="Model version created successfully",
+        )
+
+    @router.get(
+        "/models",
+        response_model=ModelVersionListResponse,
+        summary="List model versions",
+    )
+    async def list_model_versions(
+        admin_token: AdminTokenDep,
+        db: AdminDBDep,
+        status: Annotated[str | None, Query(description="Filter by status")] = None,
+        limit: Annotated[int, Query(ge=1, le=100)] = 20,
+        offset: Annotated[int, Query(ge=0)] = 0,
+    ) -> ModelVersionListResponse:
+        """List model versions with optional status filter."""
+        models, total = db.get_model_versions(status=status, limit=limit, offset=offset)
+        return ModelVersionListResponse(
+            total=total,
+            limit=limit,
+            offset=offset,
+            models=[
+                ModelVersionItem(
+                    version_id=str(m.version_id),
+                    version=m.version,
+                    name=m.name,
+                    status=m.status,
+                    is_active=m.is_active,
+                    metrics_mAP=m.metrics_mAP,
+                    document_count=m.document_count,
+                    trained_at=m.trained_at,
+                    activated_at=m.activated_at,
+                    created_at=m.created_at,
+                )
+                for m in models
+            ],
+        )
+
+    @router.get(
+        "/models/active",
+        response_model=ActiveModelResponse,
+        summary="Get active model",
+        description="Get the currently active model for inference.",
+    )
+    async def get_active_model(
+        admin_token: AdminTokenDep,
+        db: AdminDBDep,
+    ) -> ActiveModelResponse:
+        """Get the currently active model version."""
+        model = db.get_active_model_version()
+        if not model:
+            return ActiveModelResponse(has_active_model=False, model=None)
+
+        return ActiveModelResponse(
+            has_active_model=True,
+            model=ModelVersionItem(
+                version_id=str(model.version_id),
+                version=model.version,
+                name=model.name,
+                status=model.status,
+                is_active=model.is_active,
+                metrics_mAP=model.metrics_mAP,
+                document_count=model.document_count,
+                trained_at=model.trained_at,
+                activated_at=model.activated_at,
+                created_at=model.created_at,
+            ),
+        )
+
+    @router.get(
+        "/models/{version_id}",
+        response_model=ModelVersionDetailResponse,
+        summary="Get model version detail",
+    )
+    async def get_model_version(
+        version_id: str,
+        admin_token: AdminTokenDep,
+        db: AdminDBDep,
+    ) -> ModelVersionDetailResponse:
+        """Get detailed model version information."""
+        _validate_uuid(version_id, "version_id")
+        model = db.get_model_version(version_id)
+        if not model:
+            raise HTTPException(status_code=404, detail="Model version not found")
+
+        return ModelVersionDetailResponse(
+            version_id=str(model.version_id),
+            version=model.version,
+            name=model.name,
+            description=model.description,
+            model_path=model.model_path,
+            status=model.status,
+            is_active=model.is_active,
+            task_id=str(model.task_id) if model.task_id else None,
+            dataset_id=str(model.dataset_id) if model.dataset_id else None,
+            metrics_mAP=model.metrics_mAP,
+            metrics_precision=model.metrics_precision,
+            metrics_recall=model.metrics_recall,
+            document_count=model.document_count,
+            training_config=model.training_config,
+            file_size=model.file_size,
+            trained_at=model.trained_at,
+            activated_at=model.activated_at,
+            created_at=model.created_at,
+            updated_at=model.updated_at,
+        )
+
+    @router.patch(
+        "/models/{version_id}",
+        response_model=ModelVersionResponse,
+        summary="Update model version",
+    )
+    async def update_model_version(
+        version_id: str,
+        request: ModelVersionUpdateRequest,
+        admin_token: AdminTokenDep,
+        db: AdminDBDep,
+    ) -> ModelVersionResponse:
+        """Update model version metadata."""
+        _validate_uuid(version_id, "version_id")
+        model = db.update_model_version(
+            version_id=version_id,
+            name=request.name,
+            description=request.description,
+            status=request.status,
+        )
+        if not model:
+            raise HTTPException(status_code=404, detail="Model version not found")
+
+        return ModelVersionResponse(
+            version_id=str(model.version_id),
+            status=model.status,
+            message="Model version updated successfully",
+        )
+
+    @router.post(
+        "/models/{version_id}/activate",
+        response_model=ModelVersionResponse,
+        summary="Activate model version",
+        description="Activate a model version for inference (deactivates all others).",
+    )
+    async def activate_model_version(
+        version_id: str,
+        request: Request,
+        admin_token: AdminTokenDep,
+        db: AdminDBDep,
+    ) -> ModelVersionResponse:
+        """Activate a model version for inference."""
+        _validate_uuid(version_id, "version_id")
+        model = db.activate_model_version(version_id)
+        if not model:
+            raise HTTPException(status_code=404, detail="Model version not found")
+
+        # Trigger model reload in inference service
+        inference_service = getattr(request.app.state, "inference_service", None)
+        model_reloaded = False
+        if inference_service:
+            try:
+                model_reloaded = inference_service.reload_model()
+                if model_reloaded:
+                    logger.info(f"Inference model reloaded to version {model.version}")
+            except Exception as e:
+                logger.warning(f"Failed to reload inference model: {e}")
+
+        message = "Model version activated for inference"
+        if model_reloaded:
+            message += " (model reloaded)"
+
+        return ModelVersionResponse(
+            version_id=str(model.version_id),
+            status=model.status,
+            message=message,
+        )
+
+    @router.post(
+        "/models/{version_id}/deactivate",
+        response_model=ModelVersionResponse,
+        summary="Deactivate model version",
+    )
+    async def deactivate_model_version(
+        version_id: str,
+        admin_token: AdminTokenDep,
+        db: AdminDBDep,
+    ) -> ModelVersionResponse:
+        """Deactivate a model version."""
+        _validate_uuid(version_id, "version_id")
+        model = db.deactivate_model_version(version_id)
+        if not model:
+            raise HTTPException(status_code=404, detail="Model version not found")
+
+        return ModelVersionResponse(
+            version_id=str(model.version_id),
+            status=model.status,
+            message="Model version deactivated",
+        )
+
+    @router.post(
+        "/models/{version_id}/archive",
+        response_model=ModelVersionResponse,
+        summary="Archive model version",
+    )
+    async def archive_model_version(
+        version_id: str,
+        admin_token: AdminTokenDep,
+        db: AdminDBDep,
+    ) -> ModelVersionResponse:
+        """Archive a model version."""
+        _validate_uuid(version_id, "version_id")
+        model = db.archive_model_version(version_id)
+        if not model:
+            raise HTTPException(
+                status_code=400,
+                detail="Model version not found or cannot archive active model",
+            )
+
+        return ModelVersionResponse(
+            version_id=str(model.version_id),
+            status=model.status,
+            message="Model version archived",
+        )
+
+    @router.delete(
+        "/models/{version_id}",
+        summary="Delete model version",
+    )
+    async def delete_model_version(
+        version_id: str,
+        admin_token: AdminTokenDep,
+        db: AdminDBDep,
+    ) -> dict:
+        """Delete a model version."""
+        _validate_uuid(version_id, "version_id")
+        success = db.delete_model_version(version_id)
+        if not success:
+            raise HTTPException(
+                status_code=400,
+                detail="Model version not found or cannot delete active model",
+            )
+
+        return {"message": "Model version deleted"}
+
+    @router.post(
+        "/models/reload",
+        summary="Reload inference model",
+        description="Reload the inference model from the currently active model version.",
+    )
+    async def reload_inference_model(
+        request: Request,
+        admin_token: AdminTokenDep,
+    ) -> dict:
+        """Reload the inference model from active version."""
+        inference_service = getattr(request.app.state, "inference_service", None)
+        if not inference_service:
+            raise HTTPException(
+                status_code=500,
+                detail="Inference service not available",
+            )
+
+        try:
+            model_reloaded = inference_service.reload_model()
+            if model_reloaded:
+                logger.info("Inference model manually reloaded")
+                return {"message": "Model reloaded successfully", "reloaded": True}
+            else:
+                return {"message": "Model already up to date", "reloaded": False}
+        except Exception as e:
+            logger.error(f"Failed to reload model: {e}")
+            raise HTTPException(
+                status_code=500,
+                detail=f"Failed to reload model: {e}",
+            )
diff --git a/packages/inference/inference/web/app.py b/packages/inference/inference/web/app.py
index 2cfbfb6..f14e259 100644
--- a/packages/inference/inference/web/app.py
+++ b/packages/inference/inference/web/app.py
@@ -37,6 +37,7 @@ from inference.web.core.rate_limiter import RateLimiter
 # Admin API imports
 from inference.web.api.v1.admin import (
     create_annotation_router,
+    create_augmentation_router,
     create_auth_router,
     create_documents_router,
     create_locks_router,
@@ -69,10 +70,23 @@ def create_app(config: AppConfig | None = None) -> FastAPI:
     """
     config = config or default_config
 
-    # Create inference service
+    # Create model path resolver that reads from database
+    def get_active_model_path():
+        """Resolve active model path from database."""
+        try:
+            db = AdminDB()
+            active_model = db.get_active_model_version()
+            if active_model and active_model.model_path:
+                return active_model.model_path
+        except Exception as e:
+            logger.warning(f"Failed to get active model from database: {e}")
+        return None
+
+    # Create inference service with database model resolver
     inference_service = InferenceService(
         model_config=config.model,
         storage_config=config.storage,
+        model_path_resolver=get_active_model_path,
     )
 
     # Create async processing components
@@ -185,6 +199,9 @@ def create_app(config: AppConfig | None = None) -> FastAPI:
             logger.error(f"Error closing database: {e}")
 
     # Create FastAPI app
+    # Store inference service for access by routes (e.g., model reload)
+    # This will be set after app creation
+
     app = FastAPI(
         title="Invoice Field Extraction API",
         description="""
@@ -255,9 +272,15 @@ def create_app(config: AppConfig | None = None) -> FastAPI:
     training_router = create_training_router()
     app.include_router(training_router, prefix="/api/v1")
 
+    augmentation_router = create_augmentation_router()
+    app.include_router(augmentation_router, prefix="/api/v1/admin")
+
     # Include batch upload routes
     app.include_router(batch_upload_router)
 
+    # Store inference service in app state for access by routes
+    app.state.inference_service = inference_service
+
     # Root endpoint - serve HTML UI
     @app.get("/", response_class=HTMLResponse)
     async def root() -> str:
diff --git a/packages/inference/inference/web/core/scheduler.py b/packages/inference/inference/web/core/scheduler.py
index ec36469..7ece72b 100644
--- a/packages/inference/inference/web/core/scheduler.py
+++ b/packages/inference/inference/web/core/scheduler.py
@@ -110,6 +110,7 @@ class TrainingScheduler:
         try:
             # Get training configuration
             model_name = config.get("model_name", "yolo11n.pt")
+            base_model_path = config.get("base_model_path")  # For incremental training
             epochs = config.get("epochs", 100)
             batch_size = config.get("batch_size", 16)
             image_size = config.get("image_size", 640)
@@ -117,12 +118,31 @@ class TrainingScheduler:
             device = config.get("device", "0")
             project_name = config.get("project_name", "invoice_fields")
 
+            # Get augmentation config if present
+            augmentation_config = config.get("augmentation")
+            augmentation_multiplier = config.get("augmentation_multiplier", 2)
+
+            # Determine which model to use as base
+            if base_model_path:
+                # Incremental training: use existing trained model
+                if not Path(base_model_path).exists():
+                    raise ValueError(f"Base model not found: {base_model_path}")
+                effective_model = base_model_path
+                self._db.add_training_log(
+                    task_id, "INFO",
+                    f"Incremental training from: {base_model_path}",
+                )
+            else:
+                # Train from pretrained model
+                effective_model = model_name
+
             # Use dataset if available, otherwise export from scratch
             if dataset_id:
                 dataset = self._db.get_dataset(dataset_id)
                 if not dataset or not dataset.dataset_path:
                     raise ValueError(f"Dataset {dataset_id} not found or has no path")
                 data_yaml = str(Path(dataset.dataset_path) / "data.yaml")
+                dataset_path = Path(dataset.dataset_path)
                 self._db.add_training_log(
                     task_id, "INFO",
                     f"Using pre-built dataset: {dataset.name} ({dataset.total_images} images)",
@@ -132,15 +152,28 @@ class TrainingScheduler:
                 if not export_result:
                     raise ValueError("Failed to export training data")
                 data_yaml = export_result["data_yaml"]
+                dataset_path = Path(data_yaml).parent
                 self._db.add_training_log(
                     task_id, "INFO",
                     f"Exported {export_result['total_images']} images for training",
                 )
 
+            # Apply augmentation if config is provided
+            if augmentation_config and self._has_enabled_augmentations(augmentation_config):
+                aug_result = self._apply_augmentation(
+                    task_id, dataset_path, augmentation_config, augmentation_multiplier
+                )
+                if aug_result:
+                    self._db.add_training_log(
+                        task_id, "INFO",
+                        f"Augmentation complete: {aug_result['augmented_images']} new images "
+                        f"(total: {aug_result['total_images']})",
+                    )
+
             # Run YOLO training
             result = self._run_yolo_training(
                 task_id=task_id,
-                model_name=model_name,
+                model_name=effective_model,  # Use base model or pretrained model
                 data_yaml=data_yaml,
                 epochs=epochs,
                 batch_size=batch_size,
@@ -159,11 +192,94 @@ class TrainingScheduler:
             )
             self._db.add_training_log(task_id, "INFO", "Training completed successfully")
 
+            # Auto-create model version for the completed training
+            self._create_model_version_from_training(
+                task_id=task_id,
+                config=config,
+                dataset_id=dataset_id,
+                result=result,
+            )
+
         except Exception as e:
             logger.error(f"Training task {task_id} failed: {e}")
             self._db.add_training_log(task_id, "ERROR", f"Training failed: {e}")
             raise
 
+    def _create_model_version_from_training(
+        self,
+        task_id: str,
+        config: dict[str, Any],
+        dataset_id: str | None,
+        result: dict[str, Any],
+    ) -> None:
+        """Create a model version entry from completed training."""
+        try:
+            model_path = result.get("model_path")
+            if not model_path:
+                logger.warning(f"No model path in training result for task {task_id}")
+                return
+
+            # Get task info for name
+            task = self._db.get_training_task(task_id)
+            task_name = task.name if task else f"Task {task_id[:8]}"
+
+            # Generate version number based on existing versions
+            existing_versions = self._db.get_model_versions(limit=1, offset=0)
+            version_count = existing_versions[1] if existing_versions else 0
+            version = f"v{version_count + 1}.0"
+
+            # Extract metrics from result
+            metrics = result.get("metrics", {})
+            metrics_mAP = metrics.get("mAP50") or metrics.get("mAP")
+            metrics_precision = metrics.get("precision")
+            metrics_recall = metrics.get("recall")
+
+            # Get file size if possible
+            file_size = None
+            model_file = Path(model_path)
+            if model_file.exists():
+                file_size = model_file.stat().st_size
+
+            # Get document count from dataset if available
+            document_count = 0
+            if dataset_id:
+                dataset = self._db.get_dataset(dataset_id)
+                if dataset:
+                    document_count = dataset.total_documents
+
+            # Create model version
+            model_version = self._db.create_model_version(
+                version=version,
+                name=task_name,
+                model_path=str(model_path),
+                description=f"Auto-created from training task {task_id[:8]}",
+                task_id=task_id,
+                dataset_id=dataset_id,
+                metrics_mAP=metrics_mAP,
+                metrics_precision=metrics_precision,
+                metrics_recall=metrics_recall,
+                document_count=document_count,
+                training_config=config,
+                file_size=file_size,
+                trained_at=datetime.utcnow(),
+            )
+
+            logger.info(
+                f"Created model version {version} (ID: {model_version.version_id}) "
+                f"from training task {task_id}"
+            )
+            self._db.add_training_log(
+                task_id, "INFO",
+                f"Model version {version} created (mAP: {metrics_mAP:.3f if metrics_mAP else 'N/A'})",
+            )
+
+        except Exception as e:
+            logger.error(f"Failed to create model version for task {task_id}: {e}")
+            self._db.add_training_log(
+                task_id, "WARNING",
+                f"Failed to auto-create model version: {e}",
+            )
+
     def _export_training_data(self, task_id: str) -> dict[str, Any] | None:
         """Export training data for a task."""
         from pathlib import Path
@@ -256,62 +372,82 @@ names: {list(FIELD_CLASSES.values())}
         device: str,
         project_name: str,
     ) -> dict[str, Any]:
-        """Run YOLO training."""
+        """Run YOLO training using shared trainer."""
+        from shared.training import YOLOTrainer, TrainingConfig as SharedTrainingConfig
+
+        # Create log callback that writes to DB
+        def log_callback(level: str, message: str) -> None:
+            self._db.add_training_log(task_id, level, message)
+
+        # Create shared training config
+        # Note: workers=0 to avoid multiprocessing issues when running in scheduler thread
+        config = SharedTrainingConfig(
+            model_path=model_name,
+            data_yaml=data_yaml,
+            epochs=epochs,
+            batch_size=batch_size,
+            image_size=image_size,
+            learning_rate=learning_rate,
+            device=device,
+            project="runs/train",
+            name=f"{project_name}/task_{task_id[:8]}",
+            workers=0,
+        )
+
+        # Run training using shared trainer
+        trainer = YOLOTrainer(config=config, log_callback=log_callback)
+        result = trainer.train()
+
+        if not result.success:
+            raise ValueError(result.error or "Training failed")
+
+        return {
+            "model_path": result.model_path,
+            "metrics": result.metrics,
+        }
+
+    def _has_enabled_augmentations(self, aug_config: dict[str, Any]) -> bool:
+        """Check if any augmentations are enabled in the config."""
+        augmentation_fields = [
+            "perspective_warp", "wrinkle", "edge_damage", "stain",
+            "lighting_variation", "shadow", "gaussian_blur", "motion_blur",
+            "gaussian_noise", "salt_pepper", "paper_texture", "scanner_artifacts",
+        ]
+        for field in augmentation_fields:
+            if field in aug_config:
+                field_config = aug_config[field]
+                if isinstance(field_config, dict) and field_config.get("enabled", False):
+                    return True
+        return False
+
+    def _apply_augmentation(
+        self,
+        task_id: str,
+        dataset_path: Path,
+        aug_config: dict[str, Any],
+        multiplier: int,
+    ) -> dict[str, int] | None:
+        """Apply augmentation to dataset before training."""
         try:
-            from ultralytics import YOLO
-
-            # Log training start
-            self._db.add_training_log(
-                task_id, "INFO",
-                f"Starting YOLO training: model={model_name}, epochs={epochs}, batch={batch_size}",
-            )
-
-            # Load model
-            model = YOLO(model_name)
-
-            # Train
-            results = model.train(
-                data=data_yaml,
-                epochs=epochs,
-                batch=batch_size,
-                imgsz=image_size,
-                lr0=learning_rate,
-                device=device,
-                project=f"runs/train/{project_name}",
-                name=f"task_{task_id[:8]}",
-                exist_ok=True,
-                verbose=True,
-            )
-
-            # Get best model path
-            best_model = Path(results.save_dir) / "weights" / "best.pt"
-
-            # Extract metrics
-            metrics = {}
-            if hasattr(results, "results_dict"):
-                metrics = {
-                    "mAP50": results.results_dict.get("metrics/mAP50(B)", 0),
-                    "mAP50-95": results.results_dict.get("metrics/mAP50-95(B)", 0),
-                    "precision": results.results_dict.get("metrics/precision(B)", 0),
-                    "recall": results.results_dict.get("metrics/recall(B)", 0),
-                }
+            from shared.augmentation import DatasetAugmenter
 
             self._db.add_training_log(
                 task_id, "INFO",
-                f"Training completed. mAP@0.5: {metrics.get('mAP50', 'N/A')}",
+                f"Applying augmentation with multiplier={multiplier}",
             )
 
-            return {
-                "model_path": str(best_model) if best_model.exists() else None,
-                "metrics": metrics,
-            }
+            augmenter = DatasetAugmenter(aug_config)
+            result = augmenter.augment_dataset(dataset_path, multiplier=multiplier)
+
+            return result
 
-        except ImportError:
-            self._db.add_training_log(task_id, "ERROR", "Ultralytics not installed")
-            raise ValueError("Ultralytics (YOLO) not installed")
         except Exception as e:
-            self._db.add_training_log(task_id, "ERROR", f"YOLO training failed: {e}")
-            raise
+            logger.error(f"Augmentation failed for task {task_id}: {e}")
+            self._db.add_training_log(
+                task_id, "WARNING",
+                f"Augmentation failed: {e}. Continuing with original dataset.",
+            )
+            return None
 
 
 # Global scheduler instance
diff --git a/packages/inference/inference/web/schemas/admin/__init__.py b/packages/inference/inference/web/schemas/admin/__init__.py
index 1300b4e..ca4d999 100644
--- a/packages/inference/inference/web/schemas/admin/__init__.py
+++ b/packages/inference/inference/web/schemas/admin/__init__.py
@@ -10,6 +10,7 @@ from .documents import *  # noqa: F401, F403
 from .annotations import *  # noqa: F401, F403
 from .training import *  # noqa: F401, F403
 from .datasets import *  # noqa: F401, F403
+from .models import *  # noqa: F401, F403
 
 # Resolve forward references for DocumentDetailResponse
 from .documents import DocumentDetailResponse
diff --git a/packages/inference/inference/web/schemas/admin/augmentation.py b/packages/inference/inference/web/schemas/admin/augmentation.py
new file mode 100644
index 0000000..3cd638d
--- /dev/null
+++ b/packages/inference/inference/web/schemas/admin/augmentation.py
@@ -0,0 +1,187 @@
+"""Admin Augmentation Schemas."""
+
+from datetime import datetime
+from typing import Any
+
+from pydantic import BaseModel, Field
+
+
+class AugmentationParamsSchema(BaseModel):
+    """Single augmentation parameters."""
+
+    enabled: bool = Field(default=False, description="Whether this augmentation is enabled")
+    probability: float = Field(
+        default=0.5, ge=0, le=1, description="Probability of applying (0-1)"
+    )
+    params: dict[str, Any] = Field(
+        default_factory=dict, description="Type-specific parameters"
+    )
+
+
+class AugmentationConfigSchema(BaseModel):
+    """Complete augmentation configuration."""
+
+    # Geometric transforms
+    perspective_warp: AugmentationParamsSchema = Field(
+        default_factory=AugmentationParamsSchema
+    )
+
+    # Degradation effects
+    wrinkle: AugmentationParamsSchema = Field(default_factory=AugmentationParamsSchema)
+    edge_damage: AugmentationParamsSchema = Field(
+        default_factory=AugmentationParamsSchema
+    )
+    stain: AugmentationParamsSchema = Field(default_factory=AugmentationParamsSchema)
+
+    # Lighting effects
+    lighting_variation: AugmentationParamsSchema = Field(
+        default_factory=AugmentationParamsSchema
+    )
+    shadow: AugmentationParamsSchema = Field(default_factory=AugmentationParamsSchema)
+
+    # Blur effects
+    gaussian_blur: AugmentationParamsSchema = Field(
+        default_factory=AugmentationParamsSchema
+    )
+    motion_blur: AugmentationParamsSchema = Field(
+        default_factory=AugmentationParamsSchema
+    )
+
+    # Noise effects
+    gaussian_noise: AugmentationParamsSchema = Field(
+        default_factory=AugmentationParamsSchema
+    )
+    salt_pepper: AugmentationParamsSchema = Field(
+        default_factory=AugmentationParamsSchema
+    )
+
+    # Texture effects
+    paper_texture: AugmentationParamsSchema = Field(
+        default_factory=AugmentationParamsSchema
+    )
+    scanner_artifacts: AugmentationParamsSchema = Field(
+        default_factory=AugmentationParamsSchema
+    )
+
+    # Global settings
+    preserve_bboxes: bool = Field(
+        default=True, description="Whether to adjust bboxes for geometric transforms"
+    )
+    seed: int | None = Field(default=None, description="Random seed for reproducibility")
+
+
+class AugmentationTypeInfo(BaseModel):
+    """Information about an augmentation type."""
+
+    name: str = Field(..., description="Augmentation name")
+    description: str = Field(..., description="Augmentation description")
+    affects_geometry: bool = Field(
+        ..., description="Whether this augmentation affects bbox coordinates"
+    )
+    stage: str = Field(..., description="Processing stage")
+    default_params: dict[str, Any] = Field(
+        default_factory=dict, description="Default parameters"
+    )
+
+
+class AugmentationTypesResponse(BaseModel):
+    """Response for listing augmentation types."""
+
+    augmentation_types: list[AugmentationTypeInfo] = Field(
+        ..., description="Available augmentation types"
+    )
+
+
+class PresetInfo(BaseModel):
+    """Information about a preset."""
+
+    name: str = Field(..., description="Preset name")
+    description: str = Field(..., description="Preset description")
+
+
+class PresetsResponse(BaseModel):
+    """Response for listing presets."""
+
+    presets: list[PresetInfo] = Field(..., description="Available presets")
+
+
+class AugmentationPreviewRequest(BaseModel):
+    """Request to preview augmentation on an image."""
+
+    augmentation_type: str = Field(..., description="Type of augmentation to preview")
+    params: dict[str, Any] = Field(
+        default_factory=dict, description="Override parameters"
+    )
+
+
+class AugmentationPreviewResponse(BaseModel):
+    """Response with preview image data."""
+
+    preview_url: str = Field(..., description="URL to preview image")
+    original_url: str = Field(..., description="URL to original image")
+    applied_params: dict[str, Any] = Field(..., description="Applied parameters")
+
+
+class AugmentationBatchRequest(BaseModel):
+    """Request to augment a dataset offline."""
+
+    dataset_id: str = Field(..., description="Source dataset UUID")
+    config: AugmentationConfigSchema = Field(..., description="Augmentation config")
+    output_name: str = Field(
+        ..., min_length=1, max_length=255, description="Output dataset name"
+    )
+    multiplier: int = Field(
+        default=2, ge=1, le=10, description="Augmented copies per image"
+    )
+
+
+class AugmentationBatchResponse(BaseModel):
+    """Response for batch augmentation."""
+
+    task_id: str = Field(..., description="Background task UUID")
+    status: str = Field(..., description="Task status")
+    message: str = Field(..., description="Status message")
+    estimated_images: int = Field(..., description="Estimated total images")
+
+
+class AugmentedDatasetItem(BaseModel):
+    """Single augmented dataset in list."""
+
+    dataset_id: str = Field(..., description="Dataset UUID")
+    source_dataset_id: str = Field(..., description="Source dataset UUID")
+    name: str = Field(..., description="Dataset name")
+    status: str = Field(..., description="Dataset status")
+    multiplier: int = Field(..., description="Augmentation multiplier")
+    total_original_images: int = Field(..., description="Original image count")
+    total_augmented_images: int = Field(..., description="Augmented image count")
+    created_at: datetime = Field(..., description="Creation timestamp")
+
+
+class AugmentedDatasetListResponse(BaseModel):
+    """Response for listing augmented datasets."""
+
+    total: int = Field(..., ge=0, description="Total datasets")
+    limit: int = Field(..., ge=1, description="Page size")
+    offset: int = Field(..., ge=0, description="Current offset")
+    datasets: list[AugmentedDatasetItem] = Field(
+        default_factory=list, description="Dataset list"
+    )
+
+
+class AugmentedDatasetDetailResponse(BaseModel):
+    """Detailed augmented dataset response."""
+
+    dataset_id: str = Field(..., description="Dataset UUID")
+    source_dataset_id: str = Field(..., description="Source dataset UUID")
+    name: str = Field(..., description="Dataset name")
+    status: str = Field(..., description="Dataset status")
+    config: AugmentationConfigSchema | None = Field(
+        None, description="Augmentation config used"
+    )
+    multiplier: int = Field(..., description="Augmentation multiplier")
+    total_original_images: int = Field(..., description="Original image count")
+    total_augmented_images: int = Field(..., description="Augmented image count")
+    dataset_path: str | None = Field(None, description="Dataset path on disk")
+    error_message: str | None = Field(None, description="Error message if failed")
+    created_at: datetime = Field(..., description="Creation timestamp")
+    completed_at: datetime | None = Field(None, description="Completion timestamp")
diff --git a/packages/inference/inference/web/schemas/admin/datasets.py b/packages/inference/inference/web/schemas/admin/datasets.py
index f7e38c9..e1c9420 100644
--- a/packages/inference/inference/web/schemas/admin/datasets.py
+++ b/packages/inference/inference/web/schemas/admin/datasets.py
@@ -63,6 +63,8 @@ class DatasetListItem(BaseModel):
     name: str
     description: str | None
     status: str
+    training_status: str | None = None
+    active_training_task_id: str | None = None
     total_documents: int
     total_images: int
     total_annotations: int
diff --git a/packages/inference/inference/web/schemas/admin/documents.py b/packages/inference/inference/web/schemas/admin/documents.py
index fdf3874..cf3ea82 100644
--- a/packages/inference/inference/web/schemas/admin/documents.py
+++ b/packages/inference/inference/web/schemas/admin/documents.py
@@ -22,6 +22,7 @@ class DocumentUploadResponse(BaseModel):
     file_size: int = Field(..., ge=0, description="File size in bytes")
     page_count: int = Field(..., ge=1, description="Number of pages")
     status: DocumentStatus = Field(..., description="Document status")
+    group_key: str | None = Field(None, description="User-defined group key")
     auto_label_started: bool = Field(
         default=False, description="Whether auto-labeling was started"
     )
@@ -42,6 +43,7 @@ class DocumentItem(BaseModel):
     annotation_count: int = Field(default=0, ge=0, description="Number of annotations")
     upload_source: str = Field(default="ui", description="Upload source (ui or api)")
     batch_id: str | None = Field(None, description="Batch ID if uploaded via batch")
+    group_key: str | None = Field(None, description="User-defined group key")
     can_annotate: bool = Field(default=True, description="Whether document can be annotated")
     created_at: datetime = Field(..., description="Creation timestamp")
     updated_at: datetime = Field(..., description="Last update timestamp")
@@ -73,6 +75,7 @@ class DocumentDetailResponse(BaseModel):
     auto_label_error: str | None = Field(None, description="Auto-labeling error")
     upload_source: str = Field(default="ui", description="Upload source (ui or api)")
     batch_id: str | None = Field(None, description="Batch ID if uploaded via batch")
+    group_key: str | None = Field(None, description="User-defined group key")
     csv_field_values: dict[str, str] | None = Field(
         None, description="CSV field values if uploaded via batch"
     )
diff --git a/packages/inference/inference/web/schemas/admin/models.py b/packages/inference/inference/web/schemas/admin/models.py
new file mode 100644
index 0000000..7359a1e
--- /dev/null
+++ b/packages/inference/inference/web/schemas/admin/models.py
@@ -0,0 +1,95 @@
+"""Admin Model Version Schemas."""
+
+from datetime import datetime
+from typing import Any
+
+from pydantic import BaseModel, Field
+
+
+class ModelVersionCreateRequest(BaseModel):
+    """Request to create a model version."""
+
+    version: str = Field(..., min_length=1, max_length=50, description="Semantic version")
+    name: str = Field(..., min_length=1, max_length=255, description="Model name")
+    model_path: str = Field(..., min_length=1, max_length=512, description="Path to model file")
+    description: str | None = Field(None, description="Optional description")
+    task_id: str | None = Field(None, description="Training task UUID")
+    dataset_id: str | None = Field(None, description="Dataset UUID")
+    metrics_mAP: float | None = Field(None, ge=0.0, le=1.0, description="Mean Average Precision")
+    metrics_precision: float | None = Field(None, ge=0.0, le=1.0, description="Precision")
+    metrics_recall: float | None = Field(None, ge=0.0, le=1.0, description="Recall")
+    document_count: int = Field(0, ge=0, description="Documents used in training")
+    training_config: dict[str, Any] | None = Field(None, description="Training configuration")
+    file_size: int | None = Field(None, ge=0, description="Model file size in bytes")
+    trained_at: datetime | None = Field(None, description="Training completion time")
+
+
+class ModelVersionUpdateRequest(BaseModel):
+    """Request to update a model version."""
+
+    name: str | None = Field(None, min_length=1, max_length=255, description="Model name")
+    description: str | None = Field(None, description="Description")
+    status: str | None = Field(None, description="Status (inactive, archived)")
+
+
+class ModelVersionItem(BaseModel):
+    """Model version in list view."""
+
+    version_id: str = Field(..., description="Version UUID")
+    version: str = Field(..., description="Semantic version")
+    name: str = Field(..., description="Model name")
+    status: str = Field(..., description="Status (active, inactive, archived)")
+    is_active: bool = Field(..., description="Is currently active for inference")
+    metrics_mAP: float | None = Field(None, description="Mean Average Precision")
+    document_count: int = Field(..., description="Documents used in training")
+    trained_at: datetime | None = Field(None, description="Training completion time")
+    activated_at: datetime | None = Field(None, description="Last activation time")
+    created_at: datetime = Field(..., description="Creation timestamp")
+
+
+class ModelVersionListResponse(BaseModel):
+    """Paginated model version list."""
+
+    total: int = Field(..., ge=0, description="Total model versions")
+    limit: int = Field(..., ge=1, description="Page size")
+    offset: int = Field(..., ge=0, description="Current offset")
+    models: list[ModelVersionItem] = Field(default_factory=list, description="Model versions")
+
+
+class ModelVersionDetailResponse(BaseModel):
+    """Detailed model version info."""
+
+    version_id: str = Field(..., description="Version UUID")
+    version: str = Field(..., description="Semantic version")
+    name: str = Field(..., description="Model name")
+    description: str | None = Field(None, description="Description")
+    model_path: str = Field(..., description="Path to model file")
+    status: str = Field(..., description="Status (active, inactive, archived)")
+    is_active: bool = Field(..., description="Is currently active for inference")
+    task_id: str | None = Field(None, description="Training task UUID")
+    dataset_id: str | None = Field(None, description="Dataset UUID")
+    metrics_mAP: float | None = Field(None, description="Mean Average Precision")
+    metrics_precision: float | None = Field(None, description="Precision")
+    metrics_recall: float | None = Field(None, description="Recall")
+    document_count: int = Field(..., description="Documents used in training")
+    training_config: dict[str, Any] | None = Field(None, description="Training configuration")
+    file_size: int | None = Field(None, description="Model file size in bytes")
+    trained_at: datetime | None = Field(None, description="Training completion time")
+    activated_at: datetime | None = Field(None, description="Last activation time")
+    created_at: datetime = Field(..., description="Creation timestamp")
+    updated_at: datetime = Field(..., description="Last update timestamp")
+
+
+class ModelVersionResponse(BaseModel):
+    """Response for model version operation."""
+
+    version_id: str = Field(..., description="Version UUID")
+    status: str = Field(..., description="Model status")
+    message: str = Field(..., description="Status message")
+
+
+class ActiveModelResponse(BaseModel):
+    """Response for active model query."""
+
+    has_active_model: bool = Field(..., description="Whether an active model exists")
+    model: ModelVersionItem | None = Field(None, description="Active model if exists")
diff --git a/packages/inference/inference/web/schemas/admin/training.py b/packages/inference/inference/web/schemas/admin/training.py
index 6958692..2fe0cfd 100644
--- a/packages/inference/inference/web/schemas/admin/training.py
+++ b/packages/inference/inference/web/schemas/admin/training.py
@@ -5,13 +5,18 @@ from typing import Any
 
 from pydantic import BaseModel, Field
 
+from .augmentation import AugmentationConfigSchema
 from .enums import TrainingStatus, TrainingType
 
 
 class TrainingConfig(BaseModel):
     """Training configuration."""
 
-    model_name: str = Field(default="yolo11n.pt", description="Base model name")
+    model_name: str = Field(default="yolo11n.pt", description="Base model name (used if no base_model_version_id)")
+    base_model_version_id: str | None = Field(
+        default=None,
+        description="Model version UUID to use as base for incremental training. If set, uses this model instead of model_name.",
+    )
     epochs: int = Field(default=100, ge=1, le=1000, description="Training epochs")
     batch_size: int = Field(default=16, ge=1, le=128, description="Batch size")
     image_size: int = Field(default=640, ge=320, le=1280, description="Image size")
@@ -21,6 +26,18 @@ class TrainingConfig(BaseModel):
         default="invoice_fields", description="Training project name"
     )
 
+    # Data augmentation settings
+    augmentation: AugmentationConfigSchema | None = Field(
+        default=None,
+        description="Augmentation configuration. If provided, augments dataset before training.",
+    )
+    augmentation_multiplier: int = Field(
+        default=2,
+        ge=1,
+        le=10,
+        description="Number of augmented copies per original image",
+    )
+
 
 class TrainingTaskCreate(BaseModel):
     """Request to create a training task."""
diff --git a/packages/inference/inference/web/services/augmentation_service.py b/packages/inference/inference/web/services/augmentation_service.py
new file mode 100644
index 0000000..e13e22a
--- /dev/null
+++ b/packages/inference/inference/web/services/augmentation_service.py
@@ -0,0 +1,317 @@
+"""Augmentation service for handling augmentation operations."""
+
+import base64
+import io
+import re
+import uuid
+from pathlib import Path
+from typing import Any
+
+import numpy as np
+from fastapi import HTTPException
+from PIL import Image
+
+from inference.data.admin_db import AdminDB
+from inference.web.schemas.admin.augmentation import (
+    AugmentationBatchResponse,
+    AugmentationConfigSchema,
+    AugmentationPreviewResponse,
+    AugmentedDatasetItem,
+    AugmentedDatasetListResponse,
+)
+
+# Constants
+PREVIEW_MAX_SIZE = 800
+PREVIEW_SEED = 42
+UUID_PATTERN = re.compile(
+    r"^[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}$",
+    re.IGNORECASE,
+)
+
+
+class AugmentationService:
+    """Service for augmentation operations."""
+
+    def __init__(self, db: AdminDB) -> None:
+        """Initialize service with database connection."""
+        self.db = db
+
+    def _validate_uuid(self, value: str, field_name: str = "ID") -> None:
+        """
+        Validate UUID format to prevent path traversal.
+
+        Args:
+            value: Value to validate.
+            field_name: Field name for error message.
+
+        Raises:
+            HTTPException: If value is not a valid UUID.
+        """
+        if not UUID_PATTERN.match(value):
+            raise HTTPException(
+                status_code=400,
+                detail=f"Invalid {field_name} format: {value}",
+            )
+
+    async def preview_single(
+        self,
+        document_id: str,
+        page: int,
+        augmentation_type: str,
+        params: dict[str, Any],
+    ) -> AugmentationPreviewResponse:
+        """
+        Preview a single augmentation on a document page.
+
+        Args:
+            document_id: Document UUID.
+            page: Page number (1-indexed).
+            augmentation_type: Name of augmentation to apply.
+            params: Override parameters.
+
+        Returns:
+            Preview response with image URLs.
+
+        Raises:
+            HTTPException: If document not found or augmentation invalid.
+        """
+        from shared.augmentation.config import AugmentationConfig, AugmentationParams
+        from shared.augmentation.pipeline import AUGMENTATION_REGISTRY, AugmentationPipeline
+
+        # Validate augmentation type
+        if augmentation_type not in AUGMENTATION_REGISTRY:
+            raise HTTPException(
+                status_code=400,
+                detail=f"Unknown augmentation type: {augmentation_type}. "
+                f"Available: {list(AUGMENTATION_REGISTRY.keys())}",
+            )
+
+        # Get document and load image
+        image = await self._load_document_page(document_id, page)
+
+        # Create config with only this augmentation enabled
+        config_kwargs = {
+            augmentation_type: AugmentationParams(
+                enabled=True,
+                probability=1.0,  # Always apply for preview
+                params=params,
+            ),
+            "seed": PREVIEW_SEED,  # Deterministic preview
+        }
+        config = AugmentationConfig(**config_kwargs)
+        pipeline = AugmentationPipeline(config)
+
+        # Apply augmentation
+        result = pipeline.apply(image)
+
+        # Convert to base64 URLs
+        original_url = self._image_to_data_url(image)
+        preview_url = self._image_to_data_url(result.image)
+
+        return AugmentationPreviewResponse(
+            preview_url=preview_url,
+            original_url=original_url,
+            applied_params=params,
+        )
+
+    async def preview_config(
+        self,
+        document_id: str,
+        page: int,
+        config: AugmentationConfigSchema,
+    ) -> AugmentationPreviewResponse:
+        """
+        Preview full augmentation config on a document page.
+
+        Args:
+            document_id: Document UUID.
+            page: Page number (1-indexed).
+            config: Full augmentation configuration.
+
+        Returns:
+            Preview response with image URLs.
+        """
+        from shared.augmentation.config import AugmentationConfig
+        from shared.augmentation.pipeline import AugmentationPipeline
+
+        # Load image
+        image = await self._load_document_page(document_id, page)
+
+        # Convert Pydantic model to internal config
+        config_dict = config.model_dump()
+        internal_config = AugmentationConfig.from_dict(config_dict)
+        pipeline = AugmentationPipeline(internal_config)
+
+        # Apply augmentation
+        result = pipeline.apply(image)
+
+        # Convert to base64 URLs
+        original_url = self._image_to_data_url(image)
+        preview_url = self._image_to_data_url(result.image)
+
+        return AugmentationPreviewResponse(
+            preview_url=preview_url,
+            original_url=original_url,
+            applied_params=config_dict,
+        )
+
+    async def create_augmented_dataset(
+        self,
+        source_dataset_id: str,
+        config: AugmentationConfigSchema,
+        output_name: str,
+        multiplier: int,
+    ) -> AugmentationBatchResponse:
+        """
+        Create a new augmented dataset from an existing dataset.
+
+        Args:
+            source_dataset_id: Source dataset UUID.
+            config: Augmentation configuration.
+            output_name: Name for the new dataset.
+            multiplier: Number of augmented copies per image.
+
+        Returns:
+            Batch response with task ID.
+
+        Raises:
+            HTTPException: If source dataset not found.
+        """
+        # Validate source dataset exists
+        try:
+            source_dataset = self.db.get_dataset(source_dataset_id)
+            if source_dataset is None:
+                raise HTTPException(
+                    status_code=404,
+                    detail=f"Source dataset not found: {source_dataset_id}",
+                )
+        except Exception as e:
+            raise HTTPException(
+                status_code=404,
+                detail=f"Source dataset not found: {source_dataset_id}",
+            ) from e
+
+        # Create task ID for background processing
+        task_id = str(uuid.uuid4())
+
+        # Estimate total images
+        estimated_images = (
+            source_dataset.total_images * multiplier
+            if hasattr(source_dataset, "total_images")
+            else 0
+        )
+
+        # TODO: Queue background task for actual augmentation
+        # For now, return pending status
+
+        return AugmentationBatchResponse(
+            task_id=task_id,
+            status="pending",
+            message=f"Augmentation task queued for dataset '{output_name}'",
+            estimated_images=estimated_images,
+        )
+
+    async def list_augmented_datasets(
+        self,
+        limit: int = 20,
+        offset: int = 0,
+    ) -> AugmentedDatasetListResponse:
+        """
+        List augmented datasets.
+
+        Args:
+            limit: Maximum number of datasets to return.
+            offset: Number of datasets to skip.
+
+        Returns:
+            List response with datasets.
+        """
+        # TODO: Implement actual database query for augmented datasets
+        # For now, return empty list
+
+        return AugmentedDatasetListResponse(
+            total=0,
+            limit=limit,
+            offset=offset,
+            datasets=[],
+        )
+
+    async def _load_document_page(
+        self,
+        document_id: str,
+        page: int,
+    ) -> np.ndarray:
+        """
+        Load a document page as numpy array.
+
+        Args:
+            document_id: Document UUID.
+            page: Page number (1-indexed).
+
+        Returns:
+            Image as numpy array (H, W, C) with dtype uint8.
+
+        Raises:
+            HTTPException: If document or page not found.
+        """
+        # Validate document_id format to prevent path traversal
+        self._validate_uuid(document_id, "document_id")
+
+        # Get document from database
+        try:
+            document = self.db.get_document(document_id)
+            if document is None:
+                raise HTTPException(
+                    status_code=404,
+                    detail=f"Document not found: {document_id}",
+                )
+        except HTTPException:
+            raise
+        except Exception as e:
+            raise HTTPException(
+                status_code=404,
+                detail=f"Document not found: {document_id}",
+            ) from e
+
+        # Get image path for page
+        if hasattr(document, "images_dir"):
+            images_dir = Path(document.images_dir)
+        else:
+            # Fallback to constructed path
+            from inference.web.core.config import get_settings
+
+            settings = get_settings()
+            images_dir = Path(settings.admin_storage_path) / "documents" / document_id / "images"
+
+        # Find image for page
+        page_idx = page - 1  # Convert to 0-indexed
+        image_files = sorted(images_dir.glob("*.png")) + sorted(images_dir.glob("*.jpg"))
+
+        if page_idx >= len(image_files):
+            raise HTTPException(
+                status_code=404,
+                detail=f"Page {page} not found for document {document_id}",
+            )
+
+        # Load image
+        image_path = image_files[page_idx]
+        pil_image = Image.open(image_path).convert("RGB")
+        return np.array(pil_image)
+
+    def _image_to_data_url(self, image: np.ndarray) -> str:
+        """Convert numpy image to base64 data URL."""
+        pil_image = Image.fromarray(image)
+
+        # Resize for preview if too large
+        max_size = PREVIEW_MAX_SIZE
+        if max(pil_image.size) > max_size:
+            ratio = max_size / max(pil_image.size)
+            new_size = (int(pil_image.width * ratio), int(pil_image.height * ratio))
+            pil_image = pil_image.resize(new_size, Image.Resampling.LANCZOS)
+
+        # Convert to base64
+        buffer = io.BytesIO()
+        pil_image.save(buffer, format="PNG")
+        base64_data = base64.b64encode(buffer.getvalue()).decode("utf-8")
+
+        return f"data:image/png;base64,{base64_data}"
diff --git a/packages/inference/inference/web/services/dataset_builder.py b/packages/inference/inference/web/services/dataset_builder.py
index 30c69ce..fc383b4 100644
--- a/packages/inference/inference/web/services/dataset_builder.py
+++ b/packages/inference/inference/web/services/dataset_builder.py
@@ -81,29 +81,18 @@ class DatasetBuilder:
             (dataset_dir / "images" / split).mkdir(parents=True, exist_ok=True)
             (dataset_dir / "labels" / split).mkdir(parents=True, exist_ok=True)
 
-        # 3. Shuffle and split documents
+        # 3. Group documents by group_key and assign splits
         doc_list = list(documents)
-        rng = random.Random(seed)
-        rng.shuffle(doc_list)
-
-        n = len(doc_list)
-        n_train = max(1, round(n * train_ratio))
-        n_val = max(0, round(n * val_ratio))
-        n_test = n - n_train - n_val
-
-        splits = (
-            ["train"] * n_train
-            + ["val"] * n_val
-            + ["test"] * n_test
-        )
+        doc_splits = self._assign_splits_by_group(doc_list, train_ratio, val_ratio, seed)
 
         # 4. Process each document
         total_images = 0
         total_annotations = 0
         dataset_docs = []
 
-        for doc, split in zip(doc_list, splits):
+        for doc in doc_list:
             doc_id = str(doc.document_id)
+            split = doc_splits[doc_id]
             annotations = self._db.get_annotations_for_document(doc.document_id)
 
             # Group annotations by page
@@ -174,6 +163,86 @@ class DatasetBuilder:
             "total_annotations": total_annotations,
         }
 
+    def _assign_splits_by_group(
+        self,
+        documents: list,
+        train_ratio: float,
+        val_ratio: float,
+        seed: int,
+    ) -> dict[str, str]:
+        """Assign splits based on group_key.
+
+        Logic:
+        - Documents with same group_key stay together in the same split
+        - Groups with only 1 document go directly to train
+        - Groups with 2+ documents participate in shuffle & split
+
+        Args:
+            documents: List of AdminDocument objects
+            train_ratio: Fraction for training set
+            val_ratio: Fraction for validation set
+            seed: Random seed for reproducibility
+
+        Returns:
+            Dict mapping document_id (str) -> split ("train"/"val"/"test")
+        """
+        # Group documents by group_key
+        # None/empty group_key treated as unique (each doc is its own group)
+        groups: dict[str | None, list] = {}
+        for doc in documents:
+            key = doc.group_key if doc.group_key else None
+            if key is None:
+                # Treat each ungrouped doc as its own unique group
+                # Use document_id as pseudo-key
+                key = f"__ungrouped_{doc.document_id}"
+            groups.setdefault(key, []).append(doc)
+
+        # Separate single-doc groups from multi-doc groups
+        single_doc_groups: list[tuple[str | None, list]] = []
+        multi_doc_groups: list[tuple[str | None, list]] = []
+
+        for key, docs in groups.items():
+            if len(docs) == 1:
+                single_doc_groups.append((key, docs))
+            else:
+                multi_doc_groups.append((key, docs))
+
+        # Initialize result mapping
+        doc_splits: dict[str, str] = {}
+
+        # Combine all groups for splitting
+        all_groups = single_doc_groups + multi_doc_groups
+
+        # Shuffle all groups and assign splits
+        if all_groups:
+            rng = random.Random(seed)
+            rng.shuffle(all_groups)
+
+            n_groups = len(all_groups)
+            n_train = max(1, round(n_groups * train_ratio))
+            # Ensure at least 1 in val if we have more than 1 group
+            n_val = max(1 if n_groups > 1 else 0, round(n_groups * val_ratio))
+
+            for i, (_key, docs) in enumerate(all_groups):
+                if i < n_train:
+                    split = "train"
+                elif i < n_train + n_val:
+                    split = "val"
+                else:
+                    split = "test"
+
+                for doc in docs:
+                    doc_splits[str(doc.document_id)] = split
+
+        logger.info(
+            "Split assignment: %d total groups shuffled (train=%d, val=%d)",
+            len(all_groups),
+            sum(1 for s in doc_splits.values() if s == "train"),
+            sum(1 for s in doc_splits.values() if s == "val"),
+        )
+
+        return doc_splits
+
     def _generate_data_yaml(self, dataset_dir: Path) -> None:
         """Generate YOLO data.yaml configuration file."""
         data = {
diff --git a/packages/inference/inference/web/services/inference.py b/packages/inference/inference/web/services/inference.py
index f087569..84d4028 100644
--- a/packages/inference/inference/web/services/inference.py
+++ b/packages/inference/inference/web/services/inference.py
@@ -11,7 +11,7 @@ import time
 import uuid
 from dataclasses import dataclass, field
 from pathlib import Path
-from typing import TYPE_CHECKING
+from typing import TYPE_CHECKING, Callable
 
 import numpy as np
 from PIL import Image
@@ -22,6 +22,10 @@ if TYPE_CHECKING:
 logger = logging.getLogger(__name__)
 
 
+# Type alias for model path resolver function
+ModelPathResolver = Callable[[], Path | None]
+
+
 @dataclass
 class ServiceResult:
     """Result from inference service."""
@@ -42,25 +46,52 @@ class InferenceService:
     Service for running invoice field extraction.
 
     Encapsulates YOLO detection and OCR extraction logic.
+    Supports dynamic model loading from database.
     """
 
     def __init__(
         self,
         model_config: ModelConfig,
         storage_config: StorageConfig,
+        model_path_resolver: ModelPathResolver | None = None,
     ) -> None:
         """
         Initialize inference service.
 
         Args:
-            model_config: Model configuration
+            model_config: Model configuration (default model settings)
             storage_config: Storage configuration
+            model_path_resolver: Optional function to resolve model path from database.
+                                If provided, will be called to get active model path.
+                                If returns None, falls back to model_config.model_path.
         """
         self.model_config = model_config
         self.storage_config = storage_config
+        self._model_path_resolver = model_path_resolver
         self._pipeline = None
         self._detector = None
         self._is_initialized = False
+        self._current_model_path: Path | None = None
+
+    def _resolve_model_path(self) -> Path:
+        """Resolve the model path to use for inference.
+
+        Priority:
+        1. Active model from database (via resolver)
+        2. Default model from config
+        """
+        if self._model_path_resolver:
+            try:
+                db_model_path = self._model_path_resolver()
+                if db_model_path and Path(db_model_path).exists():
+                    logger.info(f"Using active model from database: {db_model_path}")
+                    return Path(db_model_path)
+                elif db_model_path:
+                    logger.warning(f"Active model path does not exist: {db_model_path}, falling back to default")
+            except Exception as e:
+                logger.warning(f"Failed to resolve model path from database: {e}, falling back to default")
+
+        return self.model_config.model_path
 
     def initialize(self) -> None:
         """Initialize the inference pipeline (lazy loading)."""
@@ -74,16 +105,20 @@ class InferenceService:
             from inference.pipeline.pipeline import InferencePipeline
             from inference.pipeline.yolo_detector import YOLODetector
 
+            # Resolve model path (from DB or config)
+            model_path = self._resolve_model_path()
+            self._current_model_path = model_path
+
             # Initialize YOLO detector for visualization
             self._detector = YOLODetector(
-                str(self.model_config.model_path),
+                str(model_path),
                 confidence_threshold=self.model_config.confidence_threshold,
                 device="cuda" if self.model_config.use_gpu else "cpu",
             )
 
             # Initialize full pipeline
             self._pipeline = InferencePipeline(
-                model_path=str(self.model_config.model_path),
+                model_path=str(model_path),
                 confidence_threshold=self.model_config.confidence_threshold,
                 use_gpu=self.model_config.use_gpu,
                 dpi=self.model_config.dpi,
@@ -92,12 +127,36 @@ class InferenceService:
 
             self._is_initialized = True
             elapsed = time.time() - start_time
-            logger.info(f"Inference service initialized in {elapsed:.2f}s")
+            logger.info(f"Inference service initialized in {elapsed:.2f}s with model: {model_path}")
 
         except Exception as e:
             logger.error(f"Failed to initialize inference service: {e}")
             raise
 
+    def reload_model(self) -> bool:
+        """Reload the model if active model has changed.
+
+        Returns:
+            True if model was reloaded, False if no change needed.
+        """
+        new_model_path = self._resolve_model_path()
+
+        if self._current_model_path == new_model_path:
+            logger.debug("Model unchanged, no reload needed")
+            return False
+
+        logger.info(f"Reloading model: {self._current_model_path} -> {new_model_path}")
+        self._is_initialized = False
+        self._pipeline = None
+        self._detector = None
+        self.initialize()
+        return True
+
+    @property
+    def current_model_path(self) -> Path | None:
+        """Get the currently loaded model path."""
+        return self._current_model_path
+
     @property
     def is_initialized(self) -> bool:
         """Check if service is initialized."""
diff --git a/packages/shared/shared/augmentation/__init__.py b/packages/shared/shared/augmentation/__init__.py
new file mode 100644
index 0000000..ec8ed5b
--- /dev/null
+++ b/packages/shared/shared/augmentation/__init__.py
@@ -0,0 +1,24 @@
+"""
+Document Image Augmentation Module.
+
+Provides augmentation transformations for training data enhancement,
+specifically designed for document images (invoices, forms, etc.).
+
+Key features:
+- Document-safe augmentations that preserve text readability
+- Support for both offline preprocessing and runtime augmentation
+- Bbox-aware geometric transforms
+- Configurable augmentation pipeline
+"""
+
+from shared.augmentation.base import AugmentationResult, BaseAugmentation
+from shared.augmentation.config import AugmentationConfig, AugmentationParams
+from shared.augmentation.dataset_augmenter import DatasetAugmenter
+
+__all__ = [
+    "AugmentationConfig",
+    "AugmentationParams",
+    "AugmentationResult",
+    "BaseAugmentation",
+    "DatasetAugmenter",
+]
diff --git a/packages/shared/shared/augmentation/base.py b/packages/shared/shared/augmentation/base.py
new file mode 100644
index 0000000..b342125
--- /dev/null
+++ b/packages/shared/shared/augmentation/base.py
@@ -0,0 +1,108 @@
+"""
+Base classes for augmentation transforms.
+
+Provides abstract base class and result dataclass for all augmentation
+implementations.
+"""
+
+from abc import ABC, abstractmethod
+from dataclasses import dataclass, field
+from typing import Any
+
+import numpy as np
+
+
+@dataclass
+class AugmentationResult:
+    """
+    Result of applying an augmentation.
+
+    Attributes:
+        image: The augmented image as numpy array (H, W, C).
+        bboxes: Updated bounding boxes if geometric transform was applied.
+                Format: (N, 5) array with [class_id, x_center, y_center, width, height].
+        transform_matrix: The transformation matrix if applicable (for bbox adjustment).
+        applied: Whether the augmentation was actually applied.
+        metadata: Additional metadata about the augmentation.
+    """
+
+    image: np.ndarray
+    bboxes: np.ndarray | None = None
+    transform_matrix: np.ndarray | None = None
+    applied: bool = True
+    metadata: dict[str, Any] | None = None
+
+
+class BaseAugmentation(ABC):
+    """
+    Abstract base class for all augmentations.
+
+    Subclasses must implement:
+        - _validate_params(): Validate augmentation parameters
+        - apply(): Apply the augmentation to an image
+
+    Class attributes:
+        name: Human-readable name of the augmentation.
+        affects_geometry: True if this augmentation modifies bbox coordinates.
+    """
+
+    name: str = "base"
+    affects_geometry: bool = False
+
+    def __init__(self, params: dict[str, Any]) -> None:
+        """
+        Initialize augmentation with parameters.
+
+        Args:
+            params: Dictionary of augmentation-specific parameters.
+        """
+        self.params = params
+        self._validate_params()
+
+    @abstractmethod
+    def _validate_params(self) -> None:
+        """
+        Validate augmentation parameters.
+
+        Raises:
+            ValueError: If parameters are invalid.
+        """
+        pass
+
+    @abstractmethod
+    def apply(
+        self,
+        image: np.ndarray,
+        bboxes: np.ndarray | None = None,
+        rng: np.random.Generator | None = None,
+    ) -> AugmentationResult:
+        """
+        Apply augmentation to image.
+
+        IMPORTANT: Implementations must NOT modify the input image or bboxes.
+        Always create copies before modifying.
+
+        Args:
+            image: Input image as numpy array (H, W, C) with dtype uint8.
+            bboxes: Optional bounding boxes in YOLO format (N, 5) array.
+                    Each row: [class_id, x_center, y_center, width, height].
+                    Coordinates are normalized to 0-1 range.
+            rng: Random number generator for reproducibility.
+                 If None, a new generator should be created.
+
+        Returns:
+            AugmentationResult with augmented image and optionally updated bboxes.
+        """
+        pass
+
+    def get_preview_params(self) -> dict[str, Any]:
+        """
+        Get parameters optimized for preview display.
+
+        Override this method to provide parameters that produce
+        clearly visible effects for preview/demo purposes.
+
+        Returns:
+            Dictionary of preview parameters.
+        """
+        return dict(self.params)
diff --git a/packages/shared/shared/augmentation/config.py b/packages/shared/shared/augmentation/config.py
new file mode 100644
index 0000000..5c7cfbc
--- /dev/null
+++ b/packages/shared/shared/augmentation/config.py
@@ -0,0 +1,274 @@
+"""
+Augmentation configuration module.
+
+Provides dataclasses for configuring document image augmentations.
+All default values are document-safe (conservative) to preserve text readability.
+"""
+
+from dataclasses import dataclass, field
+from typing import Any
+
+
+@dataclass
+class AugmentationParams:
+    """
+    Parameters for a single augmentation type.
+
+    Attributes:
+        enabled: Whether this augmentation is enabled.
+        probability: Probability of applying this augmentation (0.0 to 1.0).
+        params: Type-specific parameters dictionary.
+    """
+
+    enabled: bool = False
+    probability: float = 0.5
+    params: dict[str, Any] = field(default_factory=dict)
+
+    def to_dict(self) -> dict[str, Any]:
+        """Convert to dictionary for serialization."""
+        return {
+            "enabled": self.enabled,
+            "probability": self.probability,
+            "params": dict(self.params),
+        }
+
+    @classmethod
+    def from_dict(cls, data: dict[str, Any]) -> "AugmentationParams":
+        """Create from dictionary."""
+        return cls(
+            enabled=data.get("enabled", False),
+            probability=data.get("probability", 0.5),
+            params=dict(data.get("params", {})),
+        )
+
+
+def _default_perspective_warp() -> AugmentationParams:
+    return AugmentationParams(
+        enabled=False,
+        probability=0.3,
+        params={"max_warp": 0.02},  # Very conservative - 2% max distortion
+    )
+
+
+def _default_wrinkle() -> AugmentationParams:
+    return AugmentationParams(
+        enabled=False,
+        probability=0.3,
+        params={"intensity": 0.3, "num_wrinkles": (2, 5)},
+    )
+
+
+def _default_edge_damage() -> AugmentationParams:
+    return AugmentationParams(
+        enabled=False,
+        probability=0.2,
+        params={"max_damage_ratio": 0.05},  # Max 5% of edge damaged
+    )
+
+
+def _default_stain() -> AugmentationParams:
+    return AugmentationParams(
+        enabled=False,
+        probability=0.2,
+        params={
+            "num_stains": (1, 3),
+            "max_radius_ratio": 0.1,
+            "opacity": (0.1, 0.3),
+        },
+    )
+
+
+def _default_lighting_variation() -> AugmentationParams:
+    return AugmentationParams(
+        enabled=True,  # Safe default, commonly needed
+        probability=0.5,
+        params={
+            "brightness_range": (-0.1, 0.1),
+            "contrast_range": (0.9, 1.1),
+        },
+    )
+
+
+def _default_shadow() -> AugmentationParams:
+    return AugmentationParams(
+        enabled=False,
+        probability=0.3,
+        params={"num_shadows": (1, 2), "opacity": (0.2, 0.4)},
+    )
+
+
+def _default_gaussian_blur() -> AugmentationParams:
+    return AugmentationParams(
+        enabled=False,
+        probability=0.2,
+        params={"kernel_size": (3, 5), "sigma": (0.5, 1.5)},
+    )
+
+
+def _default_motion_blur() -> AugmentationParams:
+    return AugmentationParams(
+        enabled=False,
+        probability=0.2,
+        params={"kernel_size": (5, 9), "angle_range": (-45, 45)},
+    )
+
+
+def _default_gaussian_noise() -> AugmentationParams:
+    return AugmentationParams(
+        enabled=False,
+        probability=0.3,
+        params={"mean": 0, "std": (5, 15)},  # Conservative noise levels
+    )
+
+
+def _default_salt_pepper() -> AugmentationParams:
+    return AugmentationParams(
+        enabled=False,
+        probability=0.2,
+        params={"amount": (0.001, 0.005)},  # Very sparse
+    )
+
+
+def _default_paper_texture() -> AugmentationParams:
+    return AugmentationParams(
+        enabled=False,
+        probability=0.3,
+        params={"texture_type": "random", "intensity": (0.05, 0.15)},
+    )
+
+
+def _default_scanner_artifacts() -> AugmentationParams:
+    return AugmentationParams(
+        enabled=False,
+        probability=0.2,
+        params={"line_probability": 0.3, "dust_probability": 0.4},
+    )
+
+
+@dataclass
+class AugmentationConfig:
+    """
+    Complete augmentation configuration.
+
+    All augmentation types have document-safe defaults that preserve
+    text readability. Only lighting_variation is enabled by default.
+
+    Attributes:
+        perspective_warp: Geometric perspective transform (affects bboxes).
+        wrinkle: Paper wrinkle/crease simulation.
+        edge_damage: Damaged/torn edge effects.
+        stain: Coffee stain/smudge effects.
+        lighting_variation: Brightness and contrast variation.
+        shadow: Shadow overlay effects.
+        gaussian_blur: Gaussian blur for focus issues.
+        motion_blur: Motion blur simulation.
+        gaussian_noise: Gaussian noise for sensor noise.
+        salt_pepper: Salt and pepper noise.
+        paper_texture: Paper texture overlay.
+        scanner_artifacts: Scanner line and dust artifacts.
+        preserve_bboxes: Whether to adjust bboxes for geometric transforms.
+        seed: Random seed for reproducibility.
+    """
+
+    # Geometric transforms (affects bboxes)
+    perspective_warp: AugmentationParams = field(
+        default_factory=_default_perspective_warp
+    )
+
+    # Degradation effects
+    wrinkle: AugmentationParams = field(default_factory=_default_wrinkle)
+    edge_damage: AugmentationParams = field(default_factory=_default_edge_damage)
+    stain: AugmentationParams = field(default_factory=_default_stain)
+
+    # Lighting effects
+    lighting_variation: AugmentationParams = field(
+        default_factory=_default_lighting_variation
+    )
+    shadow: AugmentationParams = field(default_factory=_default_shadow)
+
+    # Blur effects
+    gaussian_blur: AugmentationParams = field(default_factory=_default_gaussian_blur)
+    motion_blur: AugmentationParams = field(default_factory=_default_motion_blur)
+
+    # Noise effects
+    gaussian_noise: AugmentationParams = field(default_factory=_default_gaussian_noise)
+    salt_pepper: AugmentationParams = field(default_factory=_default_salt_pepper)
+
+    # Texture effects
+    paper_texture: AugmentationParams = field(default_factory=_default_paper_texture)
+    scanner_artifacts: AugmentationParams = field(
+        default_factory=_default_scanner_artifacts
+    )
+
+    # Global settings
+    preserve_bboxes: bool = True
+    seed: int | None = None
+
+    # List of all augmentation field names
+    _AUGMENTATION_FIELDS: tuple[str, ...] = (
+        "perspective_warp",
+        "wrinkle",
+        "edge_damage",
+        "stain",
+        "lighting_variation",
+        "shadow",
+        "gaussian_blur",
+        "motion_blur",
+        "gaussian_noise",
+        "salt_pepper",
+        "paper_texture",
+        "scanner_artifacts",
+    )
+
+    def to_dict(self) -> dict[str, Any]:
+        """Convert to dictionary for serialization."""
+        result: dict[str, Any] = {
+            "preserve_bboxes": self.preserve_bboxes,
+            "seed": self.seed,
+        }
+
+        for field_name in self._AUGMENTATION_FIELDS:
+            params: AugmentationParams = getattr(self, field_name)
+            result[field_name] = params.to_dict()
+
+        return result
+
+    @classmethod
+    def from_dict(cls, data: dict[str, Any]) -> "AugmentationConfig":
+        """Create from dictionary."""
+        kwargs: dict[str, Any] = {
+            "preserve_bboxes": data.get("preserve_bboxes", True),
+            "seed": data.get("seed"),
+        }
+
+        for field_name in cls._AUGMENTATION_FIELDS:
+            if field_name in data:
+                field_data = data[field_name]
+                if isinstance(field_data, dict):
+                    kwargs[field_name] = AugmentationParams.from_dict(field_data)
+
+        return cls(**kwargs)
+
+    def get_enabled_augmentations(self) -> list[str]:
+        """Get list of enabled augmentation names."""
+        enabled = []
+        for field_name in self._AUGMENTATION_FIELDS:
+            params: AugmentationParams = getattr(self, field_name)
+            if params.enabled:
+                enabled.append(field_name)
+        return enabled
+
+    def validate(self) -> None:
+        """
+        Validate configuration.
+
+        Raises:
+            ValueError: If any configuration value is invalid.
+        """
+        for field_name in self._AUGMENTATION_FIELDS:
+            params: AugmentationParams = getattr(self, field_name)
+            if not (0.0 <= params.probability <= 1.0):
+                raise ValueError(
+                    f"{field_name}.probability must be between 0 and 1, "
+                    f"got {params.probability}"
+                )
diff --git a/packages/shared/shared/augmentation/dataset_augmenter.py b/packages/shared/shared/augmentation/dataset_augmenter.py
new file mode 100644
index 0000000..514fe71
--- /dev/null
+++ b/packages/shared/shared/augmentation/dataset_augmenter.py
@@ -0,0 +1,206 @@
+"""
+Dataset Augmenter Module.
+
+Applies augmentation pipeline to YOLO datasets,
+creating new augmented images and label files.
+"""
+
+import logging
+from pathlib import Path
+from typing import Any
+
+import numpy as np
+from PIL import Image
+
+from shared.augmentation.config import AugmentationConfig, AugmentationParams
+from shared.augmentation.pipeline import AugmentationPipeline
+
+logger = logging.getLogger(__name__)
+
+
+class DatasetAugmenter:
+    """
+    Augments YOLO datasets by creating new images and label files.
+
+    Reads images from dataset/images/train/ and labels from dataset/labels/train/,
+    applies augmentation pipeline, and saves augmented versions with "_augN" suffix.
+    """
+
+    def __init__(
+        self,
+        config: dict[str, Any],
+        seed: int | None = None,
+    ) -> None:
+        """
+        Initialize augmenter with configuration.
+
+        Args:
+            config: Dictionary mapping augmentation names to their settings.
+                    Each augmentation should have 'enabled', 'probability', and 'params'.
+            seed: Random seed for reproducibility.
+        """
+        self._config_dict = config
+        self._seed = seed
+        self._config = self._build_config(config, seed)
+
+    def _build_config(
+        self,
+        config_dict: dict[str, Any],
+        seed: int | None,
+    ) -> AugmentationConfig:
+        """Build AugmentationConfig from dictionary."""
+        kwargs: dict[str, Any] = {"seed": seed, "preserve_bboxes": True}
+
+        for aug_name, aug_settings in config_dict.items():
+            if aug_name in AugmentationConfig._AUGMENTATION_FIELDS:
+                kwargs[aug_name] = AugmentationParams(
+                    enabled=aug_settings.get("enabled", False),
+                    probability=aug_settings.get("probability", 0.5),
+                    params=aug_settings.get("params", {}),
+                )
+
+        return AugmentationConfig(**kwargs)
+
+    def augment_dataset(
+        self,
+        dataset_path: Path,
+        multiplier: int = 1,
+        split: str = "train",
+    ) -> dict[str, int]:
+        """
+        Augment a YOLO dataset.
+
+        Args:
+            dataset_path: Path to dataset root (containing images/ and labels/).
+            multiplier: Number of augmented copies per original image.
+            split: Which split to augment (default: "train").
+
+        Returns:
+            Summary dict with original_images, augmented_images, total_images.
+        """
+        images_dir = dataset_path / "images" / split
+        labels_dir = dataset_path / "labels" / split
+
+        if not images_dir.exists():
+            raise ValueError(f"Images directory not found: {images_dir}")
+
+        # Find all images
+        image_extensions = ("*.png", "*.jpg", "*.jpeg")
+        image_files: list[Path] = []
+        for ext in image_extensions:
+            image_files.extend(images_dir.glob(ext))
+
+        original_count = len(image_files)
+        augmented_count = 0
+
+        if multiplier <= 0:
+            return {
+                "original_images": original_count,
+                "augmented_images": 0,
+                "total_images": original_count,
+            }
+
+        # Process each image
+        for img_path in image_files:
+            # Load image
+            pil_image = Image.open(img_path).convert("RGB")
+            image = np.array(pil_image)
+
+            # Load corresponding label
+            label_path = labels_dir / f"{img_path.stem}.txt"
+            bboxes = self._load_bboxes(label_path) if label_path.exists() else None
+
+            # Create multiple augmented versions
+            for aug_idx in range(multiplier):
+                # Create pipeline with adjusted seed for each augmentation
+                aug_seed = None
+                if self._seed is not None:
+                    aug_seed = self._seed + aug_idx + hash(img_path.stem) % 10000
+
+                pipeline = AugmentationPipeline(
+                    self._build_config(self._config_dict, aug_seed)
+                )
+
+                # Apply augmentation
+                result = pipeline.apply(image, bboxes)
+
+                # Save augmented image
+                aug_name = f"{img_path.stem}_aug{aug_idx}{img_path.suffix}"
+                aug_img_path = images_dir / aug_name
+                aug_pil = Image.fromarray(result.image)
+                aug_pil.save(aug_img_path)
+
+                # Save augmented label
+                aug_label_path = labels_dir / f"{img_path.stem}_aug{aug_idx}.txt"
+                self._save_bboxes(aug_label_path, result.bboxes)
+
+                augmented_count += 1
+
+        logger.info(
+            "Dataset augmentation complete: %d original, %d augmented",
+            original_count,
+            augmented_count,
+        )
+
+        return {
+            "original_images": original_count,
+            "augmented_images": augmented_count,
+            "total_images": original_count + augmented_count,
+        }
+
+    def _load_bboxes(self, label_path: Path) -> np.ndarray | None:
+        """
+        Load bounding boxes from YOLO label file.
+
+        Args:
+            label_path: Path to label file.
+
+        Returns:
+            Array of shape (N, 5) with class_id, x_center, y_center, width, height.
+            Returns None if file is empty or doesn't exist.
+        """
+        if not label_path.exists():
+            return None
+
+        content = label_path.read_text().strip()
+        if not content:
+            return None
+
+        bboxes = []
+        for line in content.split("\n"):
+            parts = line.strip().split()
+            if len(parts) == 5:
+                class_id = int(parts[0])
+                x_center = float(parts[1])
+                y_center = float(parts[2])
+                width = float(parts[3])
+                height = float(parts[4])
+                bboxes.append([class_id, x_center, y_center, width, height])
+
+        if not bboxes:
+            return None
+
+        return np.array(bboxes, dtype=np.float32)
+
+    def _save_bboxes(self, label_path: Path, bboxes: np.ndarray | None) -> None:
+        """
+        Save bounding boxes to YOLO label file.
+
+        Args:
+            label_path: Path to save label file.
+            bboxes: Array of shape (N, 5) or None for empty labels.
+        """
+        if bboxes is None or len(bboxes) == 0:
+            label_path.write_text("")
+            return
+
+        lines = []
+        for bbox in bboxes:
+            class_id = int(bbox[0])
+            x_center = bbox[1]
+            y_center = bbox[2]
+            width = bbox[3]
+            height = bbox[4]
+            lines.append(f"{class_id} {x_center:.6f} {y_center:.6f} {width:.6f} {height:.6f}")
+
+        label_path.write_text("\n".join(lines))
diff --git a/packages/shared/shared/augmentation/pipeline.py b/packages/shared/shared/augmentation/pipeline.py
new file mode 100644
index 0000000..ff5bfec
--- /dev/null
+++ b/packages/shared/shared/augmentation/pipeline.py
@@ -0,0 +1,184 @@
+"""
+Augmentation pipeline module.
+
+Orchestrates multiple augmentations with proper ordering and
+provides preview functionality.
+"""
+
+from typing import Any
+
+import numpy as np
+
+from shared.augmentation.base import AugmentationResult, BaseAugmentation
+from shared.augmentation.config import AugmentationConfig, AugmentationParams
+from shared.augmentation.transforms.blur import GaussianBlur, MotionBlur
+from shared.augmentation.transforms.degradation import EdgeDamage, Stain, Wrinkle
+from shared.augmentation.transforms.geometric import PerspectiveWarp
+from shared.augmentation.transforms.lighting import LightingVariation, Shadow
+from shared.augmentation.transforms.noise import GaussianNoise, SaltPepper
+from shared.augmentation.transforms.texture import PaperTexture, ScannerArtifacts
+
+# Registry of augmentation classes
+AUGMENTATION_REGISTRY: dict[str, type[BaseAugmentation]] = {
+    "perspective_warp": PerspectiveWarp,
+    "wrinkle": Wrinkle,
+    "edge_damage": EdgeDamage,
+    "stain": Stain,
+    "lighting_variation": LightingVariation,
+    "shadow": Shadow,
+    "gaussian_blur": GaussianBlur,
+    "motion_blur": MotionBlur,
+    "gaussian_noise": GaussianNoise,
+    "salt_pepper": SaltPepper,
+    "paper_texture": PaperTexture,
+    "scanner_artifacts": ScannerArtifacts,
+}
+
+
+class AugmentationPipeline:
+    """
+    Orchestrates multiple augmentations with proper ordering.
+
+    Augmentations are applied in the following order:
+    1. Geometric (perspective_warp) - affects bboxes
+    2. Degradation (wrinkle, edge_damage, stain) - visual artifacts
+    3. Lighting (lighting_variation, shadow)
+    4. Texture (paper_texture, scanner_artifacts)
+    5. Blur (gaussian_blur, motion_blur)
+    6. Noise (gaussian_noise, salt_pepper) - applied last
+    """
+
+    STAGE_ORDER = [
+        "geometric",
+        "degradation",
+        "lighting",
+        "texture",
+        "blur",
+        "noise",
+    ]
+
+    STAGE_MAPPING = {
+        "perspective_warp": "geometric",
+        "wrinkle": "degradation",
+        "edge_damage": "degradation",
+        "stain": "degradation",
+        "lighting_variation": "lighting",
+        "shadow": "lighting",
+        "paper_texture": "texture",
+        "scanner_artifacts": "texture",
+        "gaussian_blur": "blur",
+        "motion_blur": "blur",
+        "gaussian_noise": "noise",
+        "salt_pepper": "noise",
+    }
+
+    def __init__(self, config: AugmentationConfig) -> None:
+        """
+        Initialize pipeline with configuration.
+
+        Args:
+            config: Augmentation configuration.
+        """
+        self.config = config
+        self._rng = np.random.default_rng(config.seed)
+        self._augmentations = self._build_augmentations()
+
+    def _build_augmentations(
+        self,
+    ) -> list[tuple[str, BaseAugmentation, float]]:
+        """Build ordered list of (name, augmentation, probability) tuples."""
+        augmentations: list[tuple[str, BaseAugmentation, float]] = []
+
+        for aug_name, aug_class in AUGMENTATION_REGISTRY.items():
+            params: AugmentationParams = getattr(self.config, aug_name)
+            if params.enabled:
+                aug = aug_class(params.params)
+                augmentations.append((aug_name, aug, params.probability))
+
+        # Sort by stage order
+        def sort_key(item: tuple[str, BaseAugmentation, float]) -> int:
+            name, _, _ = item
+            stage = self.STAGE_MAPPING[name]
+            return self.STAGE_ORDER.index(stage)
+
+        return sorted(augmentations, key=sort_key)
+
+    def apply(
+        self,
+        image: np.ndarray,
+        bboxes: np.ndarray | None = None,
+    ) -> AugmentationResult:
+        """
+        Apply augmentation pipeline to image.
+
+        Args:
+            image: Input image (H, W, C) as numpy array with dtype uint8.
+            bboxes: Optional bounding boxes in YOLO format (N, 5).
+
+        Returns:
+            AugmentationResult with augmented image and optionally adjusted bboxes.
+        """
+        current_image = image.copy()
+        current_bboxes = bboxes.copy() if bboxes is not None else None
+        applied_augmentations: list[str] = []
+
+        for name, aug, probability in self._augmentations:
+            if self._rng.random() < probability:
+                result = aug.apply(current_image, current_bboxes, self._rng)
+                current_image = result.image
+                if result.bboxes is not None and self.config.preserve_bboxes:
+                    current_bboxes = result.bboxes
+                applied_augmentations.append(name)
+
+        return AugmentationResult(
+            image=current_image,
+            bboxes=current_bboxes,
+            metadata={"applied_augmentations": applied_augmentations},
+        )
+
+    def preview(
+        self,
+        image: np.ndarray,
+        augmentation_name: str,
+    ) -> np.ndarray:
+        """
+        Preview a single augmentation deterministically.
+
+        Args:
+            image: Input image.
+            augmentation_name: Name of augmentation to preview.
+
+        Returns:
+            Augmented image.
+
+        Raises:
+            ValueError: If augmentation_name is not recognized.
+        """
+        if augmentation_name not in AUGMENTATION_REGISTRY:
+            raise ValueError(f"Unknown augmentation: {augmentation_name}")
+
+        params: AugmentationParams = getattr(self.config, augmentation_name)
+        aug = AUGMENTATION_REGISTRY[augmentation_name](params.params)
+
+        # Use deterministic RNG for preview
+        preview_rng = np.random.default_rng(42)
+        result = aug.apply(image.copy(), rng=preview_rng)
+        return result.image
+
+
+def get_available_augmentations() -> list[dict[str, Any]]:
+    """
+    Get list of available augmentations with metadata.
+
+    Returns:
+        List of dictionaries with augmentation info.
+    """
+    augmentations = []
+    for name, aug_class in AUGMENTATION_REGISTRY.items():
+        augmentations.append({
+            "name": name,
+            "description": aug_class.__doc__ or "",
+            "affects_geometry": aug_class.affects_geometry,
+            "stage": AugmentationPipeline.STAGE_MAPPING[name],
+        })
+    return augmentations
diff --git a/packages/shared/shared/augmentation/presets.py b/packages/shared/shared/augmentation/presets.py
new file mode 100644
index 0000000..6bad56c
--- /dev/null
+++ b/packages/shared/shared/augmentation/presets.py
@@ -0,0 +1,212 @@
+"""
+Predefined augmentation presets for common document scenarios.
+
+Presets provide ready-to-use configurations optimized for different
+use cases, from conservative (preserves text readability) to aggressive
+(simulates poor document quality).
+"""
+
+from typing import Any
+
+from shared.augmentation.config import AugmentationConfig, AugmentationParams
+
+
+PRESETS: dict[str, dict[str, Any]] = {
+    "conservative": {
+        "description": "Safe augmentations that preserve text readability",
+        "config": {
+            "lighting_variation": {
+                "enabled": True,
+                "probability": 0.5,
+                "params": {
+                    "brightness_range": (-0.1, 0.1),
+                    "contrast_range": (0.9, 1.1),
+                },
+            },
+            "gaussian_noise": {
+                "enabled": True,
+                "probability": 0.3,
+                "params": {"std": (3, 10)},
+            },
+        },
+    },
+    "moderate": {
+        "description": "Balanced augmentations for typical document degradation",
+        "config": {
+            "lighting_variation": {
+                "enabled": True,
+                "probability": 0.5,
+                "params": {
+                    "brightness_range": (-0.15, 0.15),
+                    "contrast_range": (0.85, 1.15),
+                },
+            },
+            "shadow": {
+                "enabled": True,
+                "probability": 0.3,
+                "params": {"num_shadows": (1, 2), "opacity": (0.2, 0.35)},
+            },
+            "gaussian_noise": {
+                "enabled": True,
+                "probability": 0.3,
+                "params": {"std": (5, 12)},
+            },
+            "gaussian_blur": {
+                "enabled": True,
+                "probability": 0.2,
+                "params": {"kernel_size": (3, 5), "sigma": (0.5, 1.0)},
+            },
+            "paper_texture": {
+                "enabled": True,
+                "probability": 0.3,
+                "params": {"intensity": (0.05, 0.12)},
+            },
+        },
+    },
+    "aggressive": {
+        "description": "Heavy augmentations simulating poor scan quality",
+        "config": {
+            "perspective_warp": {
+                "enabled": True,
+                "probability": 0.3,
+                "params": {"max_warp": 0.02},
+            },
+            "wrinkle": {
+                "enabled": True,
+                "probability": 0.4,
+                "params": {"intensity": 0.3, "num_wrinkles": (2, 4)},
+            },
+            "stain": {
+                "enabled": True,
+                "probability": 0.3,
+                "params": {
+                    "num_stains": (1, 2),
+                    "max_radius_ratio": 0.08,
+                    "opacity": (0.1, 0.25),
+                },
+            },
+            "lighting_variation": {
+                "enabled": True,
+                "probability": 0.6,
+                "params": {
+                    "brightness_range": (-0.2, 0.2),
+                    "contrast_range": (0.8, 1.2),
+                },
+            },
+            "shadow": {
+                "enabled": True,
+                "probability": 0.4,
+                "params": {"num_shadows": (1, 2), "opacity": (0.25, 0.4)},
+            },
+            "gaussian_blur": {
+                "enabled": True,
+                "probability": 0.3,
+                "params": {"kernel_size": (3, 5), "sigma": (0.5, 1.5)},
+            },
+            "motion_blur": {
+                "enabled": True,
+                "probability": 0.2,
+                "params": {"kernel_size": (5, 7), "angle_range": (-30, 30)},
+            },
+            "gaussian_noise": {
+                "enabled": True,
+                "probability": 0.4,
+                "params": {"std": (8, 18)},
+            },
+            "paper_texture": {
+                "enabled": True,
+                "probability": 0.4,
+                "params": {"intensity": (0.08, 0.15)},
+            },
+            "scanner_artifacts": {
+                "enabled": True,
+                "probability": 0.3,
+                "params": {"line_probability": 0.4, "dust_probability": 0.5},
+            },
+            "edge_damage": {
+                "enabled": True,
+                "probability": 0.2,
+                "params": {"max_damage_ratio": 0.04},
+            },
+        },
+    },
+    "scanned_document": {
+        "description": "Simulates typical scanned document artifacts",
+        "config": {
+            "scanner_artifacts": {
+                "enabled": True,
+                "probability": 0.5,
+                "params": {"line_probability": 0.4, "dust_probability": 0.5},
+            },
+            "paper_texture": {
+                "enabled": True,
+                "probability": 0.4,
+                "params": {"intensity": (0.05, 0.12)},
+            },
+            "lighting_variation": {
+                "enabled": True,
+                "probability": 0.3,
+                "params": {
+                    "brightness_range": (-0.1, 0.1),
+                    "contrast_range": (0.9, 1.1),
+                },
+            },
+            "gaussian_noise": {
+                "enabled": True,
+                "probability": 0.3,
+                "params": {"std": (5, 12)},
+            },
+        },
+    },
+}
+
+
+def get_preset_config(preset_name: str) -> dict[str, Any]:
+    """
+    Get the configuration dictionary for a preset.
+
+    Args:
+        preset_name: Name of the preset.
+
+    Returns:
+        Configuration dictionary.
+
+    Raises:
+        ValueError: If preset is not found.
+    """
+    if preset_name not in PRESETS:
+        raise ValueError(
+            f"Unknown preset: {preset_name}. "
+            f"Available presets: {list(PRESETS.keys())}"
+        )
+    return PRESETS[preset_name]["config"]
+
+
+def create_config_from_preset(preset_name: str) -> AugmentationConfig:
+    """
+    Create an AugmentationConfig from a preset.
+
+    Args:
+        preset_name: Name of the preset.
+
+    Returns:
+        AugmentationConfig instance.
+
+    Raises:
+        ValueError: If preset is not found.
+    """
+    config_dict = get_preset_config(preset_name)
+    return AugmentationConfig.from_dict(config_dict)
+
+
+def list_presets() -> list[dict[str, str]]:
+    """
+    List all available presets.
+
+    Returns:
+        List of dictionaries with name and description.
+    """
+    return [
+        {"name": name, "description": preset["description"]}
+        for name, preset in PRESETS.items()
+    ]
diff --git a/packages/shared/shared/augmentation/transforms/__init__.py b/packages/shared/shared/augmentation/transforms/__init__.py
new file mode 100644
index 0000000..1fe9e61
--- /dev/null
+++ b/packages/shared/shared/augmentation/transforms/__init__.py
@@ -0,0 +1,13 @@
+"""
+Augmentation transform implementations.
+
+Each module contains related augmentation classes:
+- geometric.py: Perspective warp and other geometric transforms
+- degradation.py: Wrinkle, edge damage, stain effects
+- lighting.py: Lighting variation and shadow effects
+- blur.py: Gaussian and motion blur
+- noise.py: Gaussian and salt-pepper noise
+- texture.py: Paper texture and scanner artifacts
+"""
+
+# Will be populated as transforms are implemented
diff --git a/packages/shared/shared/augmentation/transforms/blur.py b/packages/shared/shared/augmentation/transforms/blur.py
new file mode 100644
index 0000000..b5902e6
--- /dev/null
+++ b/packages/shared/shared/augmentation/transforms/blur.py
@@ -0,0 +1,144 @@
+"""
+Blur augmentation transforms.
+
+Provides blur effects for document image augmentation:
+- GaussianBlur: Simulates out-of-focus capture
+- MotionBlur: Simulates camera/document movement during capture
+"""
+
+import cv2
+import numpy as np
+
+from shared.augmentation.base import AugmentationResult, BaseAugmentation
+
+
+class GaussianBlur(BaseAugmentation):
+    """
+    Applies Gaussian blur to the image.
+
+    Simulates out-of-focus capture or low-quality optics.
+    Conservative defaults to preserve text readability.
+
+    Parameters:
+        kernel_size: Blur kernel size, int or (min, max) tuple (default: (3, 5)).
+        sigma: Blur sigma, float or (min, max) tuple (default: (0.5, 1.5)).
+    """
+
+    name = "gaussian_blur"
+    affects_geometry = False
+
+    def _validate_params(self) -> None:
+        kernel_size = self.params.get("kernel_size", (3, 5))
+        if isinstance(kernel_size, int):
+            if kernel_size < 1 or kernel_size % 2 == 0:
+                raise ValueError("kernel_size must be a positive odd integer")
+        elif isinstance(kernel_size, tuple):
+            if kernel_size[0] < 1 or kernel_size[1] < kernel_size[0]:
+                raise ValueError("kernel_size tuple must be (min, max) with min >= 1")
+
+    def apply(
+        self,
+        image: np.ndarray,
+        bboxes: np.ndarray | None = None,
+        rng: np.random.Generator | None = None,
+    ) -> AugmentationResult:
+        rng = rng or np.random.default_rng()
+
+        kernel_size = self.params.get("kernel_size", (3, 5))
+        sigma = self.params.get("sigma", (0.5, 1.5))
+
+        if isinstance(kernel_size, tuple):
+            # Choose random odd kernel size
+            min_k, max_k = kernel_size
+            possible_sizes = [k for k in range(min_k, max_k + 1) if k % 2 == 1]
+            if not possible_sizes:
+                possible_sizes = [min_k if min_k % 2 == 1 else min_k + 1]
+            kernel_size = rng.choice(possible_sizes)
+
+        if isinstance(sigma, tuple):
+            sigma = rng.uniform(sigma[0], sigma[1])
+
+        # Ensure kernel size is odd
+        if kernel_size % 2 == 0:
+            kernel_size += 1
+
+        # Apply Gaussian blur
+        blurred = cv2.GaussianBlur(image, (kernel_size, kernel_size), sigma)
+
+        return AugmentationResult(
+            image=blurred,
+            bboxes=bboxes.copy() if bboxes is not None else None,
+            metadata={"kernel_size": kernel_size, "sigma": sigma},
+        )
+
+    def get_preview_params(self) -> dict:
+        return {"kernel_size": 5, "sigma": 1.5}
+
+
+class MotionBlur(BaseAugmentation):
+    """
+    Applies motion blur to the image.
+
+    Simulates camera shake or document movement during capture.
+
+    Parameters:
+        kernel_size: Blur kernel size, int or (min, max) tuple (default: (5, 9)).
+        angle_range: Motion angle range in degrees (default: (-45, 45)).
+    """
+
+    name = "motion_blur"
+    affects_geometry = False
+
+    def _validate_params(self) -> None:
+        kernel_size = self.params.get("kernel_size", (5, 9))
+        if isinstance(kernel_size, int):
+            if kernel_size < 3:
+                raise ValueError("kernel_size must be at least 3")
+        elif isinstance(kernel_size, tuple):
+            if kernel_size[0] < 3:
+                raise ValueError("kernel_size min must be at least 3")
+
+    def apply(
+        self,
+        image: np.ndarray,
+        bboxes: np.ndarray | None = None,
+        rng: np.random.Generator | None = None,
+    ) -> AugmentationResult:
+        rng = rng or np.random.default_rng()
+
+        kernel_size = self.params.get("kernel_size", (5, 9))
+        angle_range = self.params.get("angle_range", (-45, 45))
+
+        if isinstance(kernel_size, tuple):
+            kernel_size = rng.integers(kernel_size[0], kernel_size[1] + 1)
+
+        angle = rng.uniform(angle_range[0], angle_range[1])
+
+        # Create motion blur kernel
+        kernel = np.zeros((kernel_size, kernel_size), dtype=np.float32)
+
+        # Draw a line in the center of the kernel
+        center = kernel_size // 2
+        angle_rad = np.deg2rad(angle)
+
+        for i in range(kernel_size):
+            offset = i - center
+            x = int(center + offset * np.cos(angle_rad))
+            y = int(center + offset * np.sin(angle_rad))
+            if 0 <= x < kernel_size and 0 <= y < kernel_size:
+                kernel[y, x] = 1.0
+
+        # Normalize kernel
+        kernel = kernel / kernel.sum() if kernel.sum() > 0 else kernel
+
+        # Apply motion blur
+        blurred = cv2.filter2D(image, -1, kernel)
+
+        return AugmentationResult(
+            image=blurred,
+            bboxes=bboxes.copy() if bboxes is not None else None,
+            metadata={"kernel_size": kernel_size, "angle": angle},
+        )
+
+    def get_preview_params(self) -> dict:
+        return {"kernel_size": 7, "angle_range": (-30, 30)}
diff --git a/packages/shared/shared/augmentation/transforms/degradation.py b/packages/shared/shared/augmentation/transforms/degradation.py
new file mode 100644
index 0000000..b5c4560
--- /dev/null
+++ b/packages/shared/shared/augmentation/transforms/degradation.py
@@ -0,0 +1,259 @@
+"""
+Degradation augmentation transforms.
+
+Provides degradation effects for document image augmentation:
+- Wrinkle: Paper wrinkle/crease simulation
+- EdgeDamage: Damaged/torn edge effects
+- Stain: Coffee stain/smudge effects
+"""
+
+import cv2
+import numpy as np
+
+from shared.augmentation.base import AugmentationResult, BaseAugmentation
+
+
+class Wrinkle(BaseAugmentation):
+    """
+    Simulates paper wrinkles/creases using displacement mapping.
+
+    Document-friendly: Uses subtle displacement to preserve text readability.
+
+    Parameters:
+        intensity: Wrinkle intensity (0-1) (default: 0.3).
+        num_wrinkles: Number of wrinkles, int or (min, max) tuple (default: (2, 5)).
+    """
+
+    name = "wrinkle"
+    affects_geometry = False
+
+    def _validate_params(self) -> None:
+        intensity = self.params.get("intensity", 0.3)
+        if not (0 < intensity <= 1):
+            raise ValueError("intensity must be between 0 and 1")
+
+    def apply(
+        self,
+        image: np.ndarray,
+        bboxes: np.ndarray | None = None,
+        rng: np.random.Generator | None = None,
+    ) -> AugmentationResult:
+        rng = rng or np.random.default_rng()
+
+        h, w = image.shape[:2]
+        intensity = self.params.get("intensity", 0.3)
+        num_wrinkles = self.params.get("num_wrinkles", (2, 5))
+
+        if isinstance(num_wrinkles, tuple):
+            num_wrinkles = rng.integers(num_wrinkles[0], num_wrinkles[1] + 1)
+
+        # Create displacement maps
+        displacement_x = np.zeros((h, w), dtype=np.float32)
+        displacement_y = np.zeros((h, w), dtype=np.float32)
+
+        for _ in range(num_wrinkles):
+            # Random wrinkle parameters
+            angle = rng.uniform(0, np.pi)
+            x0 = rng.uniform(0, w)
+            y0 = rng.uniform(0, h)
+            length = rng.uniform(0.3, 0.8) * min(h, w)
+            width = rng.uniform(0.02, 0.05) * min(h, w)
+
+            # Create coordinate grids
+            xx, yy = np.meshgrid(np.arange(w), np.arange(h))
+
+            # Distance from wrinkle line
+            dx = (xx - x0) * np.cos(angle) + (yy - y0) * np.sin(angle)
+            dy = -(xx - x0) * np.sin(angle) + (yy - y0) * np.cos(angle)
+
+            # Gaussian falloff perpendicular to wrinkle
+            mask = np.exp(-dy**2 / (2 * width**2))
+            mask *= (np.abs(dx) < length / 2).astype(np.float32)
+
+            # Displacement perpendicular to wrinkle
+            disp_amount = intensity * rng.uniform(2, 8)
+            displacement_x += mask * disp_amount * np.sin(angle)
+            displacement_y += mask * disp_amount * np.cos(angle)
+
+        # Create remap coordinates
+        map_x = (np.arange(w)[np.newaxis, :] + displacement_x).astype(np.float32)
+        map_y = (np.arange(h)[:, np.newaxis] + displacement_y).astype(np.float32)
+
+        # Apply displacement
+        augmented = cv2.remap(
+            image, map_x, map_y, cv2.INTER_LINEAR, borderMode=cv2.BORDER_REFLECT
+        )
+
+        # Add subtle shading along wrinkles
+        max_disp = np.max(np.abs(displacement_y)) + 1e-6
+        shading = 1 - 0.1 * intensity * np.abs(displacement_y) / max_disp
+        shading = shading[:, :, np.newaxis]
+        augmented = (augmented.astype(np.float32) * shading).astype(np.uint8)
+
+        return AugmentationResult(
+            image=augmented,
+            bboxes=bboxes.copy() if bboxes is not None else None,
+            metadata={"num_wrinkles": num_wrinkles, "intensity": intensity},
+        )
+
+    def get_preview_params(self) -> dict:
+        return {"intensity": 0.5, "num_wrinkles": 3}
+
+
+class EdgeDamage(BaseAugmentation):
+    """
+    Adds damaged/torn edge effects to the image.
+
+    Simulates worn or torn document edges.
+
+    Parameters:
+        max_damage_ratio: Maximum proportion of edge to damage (default: 0.05).
+        edges: Which edges to potentially damage (default: all).
+    """
+
+    name = "edge_damage"
+    affects_geometry = False
+
+    def _validate_params(self) -> None:
+        max_damage_ratio = self.params.get("max_damage_ratio", 0.05)
+        if not (0 < max_damage_ratio <= 0.2):
+            raise ValueError("max_damage_ratio must be between 0 and 0.2")
+
+    def apply(
+        self,
+        image: np.ndarray,
+        bboxes: np.ndarray | None = None,
+        rng: np.random.Generator | None = None,
+    ) -> AugmentationResult:
+        rng = rng or np.random.default_rng()
+
+        h, w = image.shape[:2]
+        max_damage_ratio = self.params.get("max_damage_ratio", 0.05)
+        edges = self.params.get("edges", ["top", "bottom", "left", "right"])
+
+        output = image.copy()
+
+        # Select random edge to damage
+        edge = rng.choice(edges)
+        damage_size = int(max_damage_ratio * min(h, w))
+
+        if edge == "top":
+            # Create irregular top edge
+            for x in range(w):
+                depth = rng.integers(0, damage_size + 1)
+                if depth > 0:
+                    # Random color (white or darker)
+                    color = rng.integers(200, 255) if rng.random() > 0.5 else rng.integers(100, 150)
+                    output[:depth, x] = color
+
+        elif edge == "bottom":
+            for x in range(w):
+                depth = rng.integers(0, damage_size + 1)
+                if depth > 0:
+                    color = rng.integers(200, 255) if rng.random() > 0.5 else rng.integers(100, 150)
+                    output[h - depth:, x] = color
+
+        elif edge == "left":
+            for y in range(h):
+                depth = rng.integers(0, damage_size + 1)
+                if depth > 0:
+                    color = rng.integers(200, 255) if rng.random() > 0.5 else rng.integers(100, 150)
+                    output[y, :depth] = color
+
+        else:  # right
+            for y in range(h):
+                depth = rng.integers(0, damage_size + 1)
+                if depth > 0:
+                    color = rng.integers(200, 255) if rng.random() > 0.5 else rng.integers(100, 150)
+                    output[y, w - depth:] = color
+
+        return AugmentationResult(
+            image=output,
+            bboxes=bboxes.copy() if bboxes is not None else None,
+            metadata={"edge": edge, "damage_size": damage_size},
+        )
+
+    def get_preview_params(self) -> dict:
+        return {"max_damage_ratio": 0.08}
+
+
+class Stain(BaseAugmentation):
+    """
+    Adds coffee stain/smudge effects to the image.
+
+    Simulates accidental stains on documents.
+
+    Parameters:
+        num_stains: Number of stains, int or (min, max) tuple (default: (1, 3)).
+        max_radius_ratio: Maximum stain radius as ratio of image size (default: 0.1).
+        opacity: Stain opacity, float or (min, max) tuple (default: (0.1, 0.3)).
+    """
+
+    name = "stain"
+    affects_geometry = False
+
+    def _validate_params(self) -> None:
+        opacity = self.params.get("opacity", (0.1, 0.3))
+        if isinstance(opacity, (int, float)):
+            if not (0 < opacity <= 1):
+                raise ValueError("opacity must be between 0 and 1")
+
+    def apply(
+        self,
+        image: np.ndarray,
+        bboxes: np.ndarray | None = None,
+        rng: np.random.Generator | None = None,
+    ) -> AugmentationResult:
+        rng = rng or np.random.default_rng()
+
+        h, w = image.shape[:2]
+        num_stains = self.params.get("num_stains", (1, 3))
+        max_radius_ratio = self.params.get("max_radius_ratio", 0.1)
+        opacity = self.params.get("opacity", (0.1, 0.3))
+
+        if isinstance(num_stains, tuple):
+            num_stains = rng.integers(num_stains[0], num_stains[1] + 1)
+        if isinstance(opacity, tuple):
+            opacity = rng.uniform(opacity[0], opacity[1])
+
+        output = image.astype(np.float32)
+        max_radius = int(max_radius_ratio * min(h, w))
+
+        for _ in range(num_stains):
+            # Random stain position and size
+            cx = rng.integers(max_radius, w - max_radius)
+            cy = rng.integers(max_radius, h - max_radius)
+            radius = rng.integers(max_radius // 3, max_radius)
+
+            # Create stain mask with irregular edges
+            yy, xx = np.ogrid[:h, :w]
+            dist = np.sqrt((xx - cx) ** 2 + (yy - cy) ** 2)
+
+            # Add noise to make edges irregular
+            noise = rng.uniform(0.8, 1.2, (h, w))
+            mask = (dist < radius * noise).astype(np.float32)
+
+            # Blur for soft edges
+            mask = cv2.GaussianBlur(mask, (21, 21), 0)
+
+            # Random stain color (brownish/yellowish)
+            stain_color = np.array([
+                rng.integers(180, 220),  # R
+                rng.integers(160, 200),  # G
+                rng.integers(120, 160),  # B
+            ], dtype=np.float32)
+
+            # Apply stain
+            mask_3d = mask[:, :, np.newaxis]
+            output = output * (1 - mask_3d * opacity) + stain_color * mask_3d * opacity
+
+        output = np.clip(output, 0, 255).astype(np.uint8)
+
+        return AugmentationResult(
+            image=output,
+            bboxes=bboxes.copy() if bboxes is not None else None,
+            metadata={"num_stains": num_stains, "opacity": opacity},
+        )
+
+    def get_preview_params(self) -> dict:
+        return {"num_stains": 2, "max_radius_ratio": 0.1, "opacity": 0.25}
diff --git a/packages/shared/shared/augmentation/transforms/geometric.py b/packages/shared/shared/augmentation/transforms/geometric.py
new file mode 100644
index 0000000..04dc0be
--- /dev/null
+++ b/packages/shared/shared/augmentation/transforms/geometric.py
@@ -0,0 +1,145 @@
+"""
+Geometric augmentation transforms.
+
+Provides geometric transforms for document image augmentation:
+- PerspectiveWarp: Subtle perspective distortion
+"""
+
+import cv2
+import numpy as np
+
+from shared.augmentation.base import AugmentationResult, BaseAugmentation
+
+
+class PerspectiveWarp(BaseAugmentation):
+    """
+    Applies subtle perspective transformation to the image.
+
+    Simulates viewing document at slight angle. Very conservative
+    by default to preserve text readability.
+
+    IMPORTANT: This transform affects bounding box coordinates.
+
+    Parameters:
+        max_warp: Maximum warp as proportion of image size (default: 0.02).
+    """
+
+    name = "perspective_warp"
+    affects_geometry = True
+
+    def _validate_params(self) -> None:
+        max_warp = self.params.get("max_warp", 0.02)
+        if not (0 < max_warp <= 0.1):
+            raise ValueError("max_warp must be between 0 and 0.1")
+
+    def apply(
+        self,
+        image: np.ndarray,
+        bboxes: np.ndarray | None = None,
+        rng: np.random.Generator | None = None,
+    ) -> AugmentationResult:
+        rng = rng or np.random.default_rng()
+
+        h, w = image.shape[:2]
+        max_warp = self.params.get("max_warp", 0.02)
+
+        # Original corners
+        src_pts = np.float32([
+            [0, 0],
+            [w, 0],
+            [w, h],
+            [0, h],
+        ])
+
+        # Add random perturbations to corners
+        max_offset = max_warp * min(h, w)
+        dst_pts = src_pts.copy()
+        for i in range(4):
+            dst_pts[i, 0] += rng.uniform(-max_offset, max_offset)
+            dst_pts[i, 1] += rng.uniform(-max_offset, max_offset)
+
+        # Compute perspective transform matrix
+        transform_matrix = cv2.getPerspectiveTransform(src_pts, dst_pts)
+
+        # Apply perspective transform
+        warped = cv2.warpPerspective(
+            image, transform_matrix, (w, h),
+            borderMode=cv2.BORDER_REPLICATE
+        )
+
+        # Transform bounding boxes if present
+        transformed_bboxes = None
+        if bboxes is not None:
+            transformed_bboxes = self._transform_bboxes(
+                bboxes, transform_matrix, w, h
+            )
+
+        return AugmentationResult(
+            image=warped,
+            bboxes=transformed_bboxes,
+            transform_matrix=transform_matrix,
+            metadata={"max_warp": max_warp},
+        )
+
+    def _transform_bboxes(
+        self,
+        bboxes: np.ndarray,
+        transform_matrix: np.ndarray,
+        w: int,
+        h: int,
+    ) -> np.ndarray:
+        """Transform bounding boxes using perspective matrix."""
+        if len(bboxes) == 0:
+            return bboxes.copy()
+
+        transformed = []
+        for bbox in bboxes:
+            class_id, x_center, y_center, width, height = bbox
+
+            # Convert normalized coords to pixel coords
+            x_center_px = x_center * w
+            y_center_px = y_center * h
+            width_px = width * w
+            height_px = height * h
+
+            # Get corner points
+            x1 = x_center_px - width_px / 2
+            y1 = y_center_px - height_px / 2
+            x2 = x_center_px + width_px / 2
+            y2 = y_center_px + height_px / 2
+
+            # Transform all 4 corners
+            corners = np.float32([
+                [x1, y1],
+                [x2, y1],
+                [x2, y2],
+                [x1, y2],
+            ]).reshape(-1, 1, 2)
+
+            transformed_corners = cv2.perspectiveTransform(corners, transform_matrix)
+            transformed_corners = transformed_corners.reshape(-1, 2)
+
+            # Get bounding box of transformed corners
+            new_x1 = np.min(transformed_corners[:, 0])
+            new_y1 = np.min(transformed_corners[:, 1])
+            new_x2 = np.max(transformed_corners[:, 0])
+            new_y2 = np.max(transformed_corners[:, 1])
+
+            # Convert back to normalized center format
+            new_width = (new_x2 - new_x1) / w
+            new_height = (new_y2 - new_y1) / h
+            new_x_center = ((new_x1 + new_x2) / 2) / w
+            new_y_center = ((new_y1 + new_y2) / 2) / h
+
+            # Clamp to valid range
+            new_x_center = np.clip(new_x_center, 0, 1)
+            new_y_center = np.clip(new_y_center, 0, 1)
+            new_width = np.clip(new_width, 0, 1)
+            new_height = np.clip(new_height, 0, 1)
+
+            transformed.append([class_id, new_x_center, new_y_center, new_width, new_height])
+
+        return np.array(transformed, dtype=np.float32)
+
+    def get_preview_params(self) -> dict:
+        return {"max_warp": 0.03}
diff --git a/packages/shared/shared/augmentation/transforms/lighting.py b/packages/shared/shared/augmentation/transforms/lighting.py
new file mode 100644
index 0000000..93341c9
--- /dev/null
+++ b/packages/shared/shared/augmentation/transforms/lighting.py
@@ -0,0 +1,167 @@
+"""
+Lighting augmentation transforms.
+
+Provides lighting effects for document image augmentation:
+- LightingVariation: Adjusts brightness and contrast
+- Shadow: Adds shadow overlay effects
+"""
+
+import cv2
+import numpy as np
+
+from shared.augmentation.base import AugmentationResult, BaseAugmentation
+
+
+class LightingVariation(BaseAugmentation):
+    """
+    Adjusts image brightness and contrast.
+
+    Simulates different lighting conditions during document capture.
+    Safe for documents with conservative default parameters.
+
+    Parameters:
+        brightness_range: (min, max) brightness adjustment (default: (-0.1, 0.1)).
+        contrast_range: (min, max) contrast multiplier (default: (0.9, 1.1)).
+    """
+
+    name = "lighting_variation"
+    affects_geometry = False
+
+    def _validate_params(self) -> None:
+        brightness = self.params.get("brightness_range", (-0.1, 0.1))
+        contrast = self.params.get("contrast_range", (0.9, 1.1))
+
+        if not isinstance(brightness, tuple) or len(brightness) != 2:
+            raise ValueError("brightness_range must be a (min, max) tuple")
+        if not isinstance(contrast, tuple) or len(contrast) != 2:
+            raise ValueError("contrast_range must be a (min, max) tuple")
+
+    def apply(
+        self,
+        image: np.ndarray,
+        bboxes: np.ndarray | None = None,
+        rng: np.random.Generator | None = None,
+    ) -> AugmentationResult:
+        rng = rng or np.random.default_rng()
+
+        brightness_range = self.params.get("brightness_range", (-0.1, 0.1))
+        contrast_range = self.params.get("contrast_range", (0.9, 1.1))
+
+        # Random brightness and contrast
+        brightness = rng.uniform(brightness_range[0], brightness_range[1])
+        contrast = rng.uniform(contrast_range[0], contrast_range[1])
+
+        # Apply adjustments
+        adjusted = image.astype(np.float32)
+
+        # Contrast adjustment (multiply around mean)
+        mean = adjusted.mean()
+        adjusted = (adjusted - mean) * contrast + mean
+
+        # Brightness adjustment (add offset)
+        adjusted = adjusted + brightness * 255
+
+        # Clip and convert back
+        adjusted = np.clip(adjusted, 0, 255).astype(np.uint8)
+
+        return AugmentationResult(
+            image=adjusted,
+            bboxes=bboxes.copy() if bboxes is not None else None,
+            metadata={"brightness": brightness, "contrast": contrast},
+        )
+
+    def get_preview_params(self) -> dict:
+        return {"brightness_range": (-0.15, 0.15), "contrast_range": (0.85, 1.15)}
+
+
+class Shadow(BaseAugmentation):
+    """
+    Adds shadow overlay effects to the image.
+
+    Simulates shadows from objects or hands during document capture.
+
+    Parameters:
+        num_shadows: Number of shadow regions, int or (min, max) tuple (default: (1, 2)).
+        opacity: Shadow darkness, float or (min, max) tuple (default: (0.2, 0.4)).
+    """
+
+    name = "shadow"
+    affects_geometry = False
+
+    def _validate_params(self) -> None:
+        opacity = self.params.get("opacity", (0.2, 0.4))
+        if isinstance(opacity, (int, float)):
+            if not (0 <= opacity <= 1):
+                raise ValueError("opacity must be between 0 and 1")
+        elif isinstance(opacity, tuple):
+            if not (0 <= opacity[0] <= opacity[1] <= 1):
+                raise ValueError("opacity tuple must be in range [0, 1]")
+
+    def apply(
+        self,
+        image: np.ndarray,
+        bboxes: np.ndarray | None = None,
+        rng: np.random.Generator | None = None,
+    ) -> AugmentationResult:
+        rng = rng or np.random.default_rng()
+
+        num_shadows = self.params.get("num_shadows", (1, 2))
+        opacity = self.params.get("opacity", (0.2, 0.4))
+
+        if isinstance(num_shadows, tuple):
+            num_shadows = rng.integers(num_shadows[0], num_shadows[1] + 1)
+        if isinstance(opacity, tuple):
+            opacity = rng.uniform(opacity[0], opacity[1])
+
+        h, w = image.shape[:2]
+        output = image.astype(np.float32)
+
+        for _ in range(num_shadows):
+            # Generate random shadow polygon
+            num_vertices = rng.integers(3, 6)
+            vertices = []
+
+            # Start from a random edge
+            edge = rng.integers(0, 4)
+            if edge == 0:  # Top
+                start = (rng.integers(0, w), 0)
+            elif edge == 1:  # Right
+                start = (w, rng.integers(0, h))
+            elif edge == 2:  # Bottom
+                start = (rng.integers(0, w), h)
+            else:  # Left
+                start = (0, rng.integers(0, h))
+
+            vertices.append(start)
+
+            # Add random vertices
+            for _ in range(num_vertices - 1):
+                x = rng.integers(0, w)
+                y = rng.integers(0, h)
+                vertices.append((x, y))
+
+            # Create shadow mask
+            mask = np.zeros((h, w), dtype=np.float32)
+            pts = np.array(vertices, dtype=np.int32).reshape((-1, 1, 2))
+            cv2.fillPoly(mask, [pts], 1.0)
+
+            # Blur the mask for soft edges
+            blur_size = max(31, min(h, w) // 10)
+            if blur_size % 2 == 0:
+                blur_size += 1
+            mask = cv2.GaussianBlur(mask, (blur_size, blur_size), 0)
+
+            # Apply shadow
+            shadow_factor = 1 - opacity * mask[:, :, np.newaxis]
+            output = output * shadow_factor
+
+        output = np.clip(output, 0, 255).astype(np.uint8)
+
+        return AugmentationResult(
+            image=output,
+            bboxes=bboxes.copy() if bboxes is not None else None,
+            metadata={"num_shadows": num_shadows, "opacity": opacity},
+        )
+
+    def get_preview_params(self) -> dict:
+        return {"num_shadows": 1, "opacity": 0.3}
diff --git a/packages/shared/shared/augmentation/transforms/noise.py b/packages/shared/shared/augmentation/transforms/noise.py
new file mode 100644
index 0000000..c0b6eb2
--- /dev/null
+++ b/packages/shared/shared/augmentation/transforms/noise.py
@@ -0,0 +1,142 @@
+"""
+Noise augmentation transforms.
+
+Provides noise effects for document image augmentation:
+- GaussianNoise: Adds Gaussian noise to simulate sensor noise
+- SaltPepper: Adds salt and pepper noise for impulse noise effects
+"""
+
+from typing import Any
+
+import numpy as np
+
+from shared.augmentation.base import AugmentationResult, BaseAugmentation
+
+
+class GaussianNoise(BaseAugmentation):
+    """
+    Adds Gaussian noise to the image.
+
+    Simulates sensor noise from cameras or scanners.
+    Document-safe with conservative default parameters.
+
+    Parameters:
+        mean: Mean of the Gaussian noise (default: 0).
+        std: Standard deviation, can be int or (min, max) tuple (default: (5, 15)).
+    """
+
+    name = "gaussian_noise"
+    affects_geometry = False
+
+    def _validate_params(self) -> None:
+        std = self.params.get("std", (5, 15))
+        if isinstance(std, (int, float)):
+            if std < 0:
+                raise ValueError("std must be non-negative")
+        elif isinstance(std, tuple):
+            if len(std) != 2 or std[0] < 0 or std[1] < std[0]:
+                raise ValueError("std tuple must be (min, max) with min <= max >= 0")
+
+    def apply(
+        self,
+        image: np.ndarray,
+        bboxes: np.ndarray | None = None,
+        rng: np.random.Generator | None = None,
+    ) -> AugmentationResult:
+        rng = rng or np.random.default_rng()
+
+        mean = self.params.get("mean", 0)
+        std = self.params.get("std", (5, 15))
+
+        if isinstance(std, tuple):
+            std = rng.uniform(std[0], std[1])
+
+        # Generate noise
+        noise = rng.normal(mean, std, image.shape).astype(np.float32)
+
+        # Apply noise
+        noisy = image.astype(np.float32) + noise
+        noisy = np.clip(noisy, 0, 255).astype(np.uint8)
+
+        return AugmentationResult(
+            image=noisy,
+            bboxes=bboxes.copy() if bboxes is not None else None,
+            metadata={"applied_std": std},
+        )
+
+    def get_preview_params(self) -> dict[str, Any]:
+        return {"mean": 0, "std": 15}
+
+
+class SaltPepper(BaseAugmentation):
+    """
+    Adds salt and pepper (impulse) noise to the image.
+
+    Simulates defects from damaged sensors or transmission errors.
+    Very sparse by default to preserve document readability.
+
+    Parameters:
+        amount: Proportion of pixels to affect, can be float or (min, max) tuple.
+                Default: (0.001, 0.005) for very sparse noise.
+        salt_vs_pepper: Ratio of salt to pepper (default: 0.5 for equal amounts).
+    """
+
+    name = "salt_pepper"
+    affects_geometry = False
+
+    def _validate_params(self) -> None:
+        amount = self.params.get("amount", (0.001, 0.005))
+        if isinstance(amount, (int, float)):
+            if not (0 <= amount <= 1):
+                raise ValueError("amount must be between 0 and 1")
+        elif isinstance(amount, tuple):
+            if len(amount) != 2 or not (0 <= amount[0] <= amount[1] <= 1):
+                raise ValueError("amount tuple must be (min, max) in range [0, 1]")
+
+    def apply(
+        self,
+        image: np.ndarray,
+        bboxes: np.ndarray | None = None,
+        rng: np.random.Generator | None = None,
+    ) -> AugmentationResult:
+        rng = rng or np.random.default_rng()
+
+        amount = self.params.get("amount", (0.001, 0.005))
+        salt_vs_pepper = self.params.get("salt_vs_pepper", 0.5)
+
+        if isinstance(amount, tuple):
+            amount = rng.uniform(amount[0], amount[1])
+
+        # Copy image
+        output = image.copy()
+        h, w = image.shape[:2]
+        total_pixels = h * w
+
+        # Calculate number of salt and pepper pixels
+        num_salt = int(total_pixels * amount * salt_vs_pepper)
+        num_pepper = int(total_pixels * amount * (1 - salt_vs_pepper))
+
+        # Add salt (white pixels)
+        if num_salt > 0:
+            salt_coords = (
+                rng.integers(0, h, num_salt),
+                rng.integers(0, w, num_salt),
+            )
+            output[salt_coords] = 255
+
+        # Add pepper (black pixels)
+        if num_pepper > 0:
+            pepper_coords = (
+                rng.integers(0, h, num_pepper),
+                rng.integers(0, w, num_pepper),
+            )
+            output[pepper_coords] = 0
+
+        return AugmentationResult(
+            image=output,
+            bboxes=bboxes.copy() if bboxes is not None else None,
+            metadata={"applied_amount": amount},
+        )
+
+    def get_preview_params(self) -> dict[str, Any]:
+        return {"amount": 0.01, "salt_vs_pepper": 0.5}
diff --git a/packages/shared/shared/augmentation/transforms/texture.py b/packages/shared/shared/augmentation/transforms/texture.py
new file mode 100644
index 0000000..41287d2
--- /dev/null
+++ b/packages/shared/shared/augmentation/transforms/texture.py
@@ -0,0 +1,159 @@
+"""
+Texture augmentation transforms.
+
+Provides texture effects for document image augmentation:
+- PaperTexture: Adds paper grain/texture
+- ScannerArtifacts: Adds scanner line and dust artifacts
+"""
+
+import cv2
+import numpy as np
+
+from shared.augmentation.base import AugmentationResult, BaseAugmentation
+
+
+class PaperTexture(BaseAugmentation):
+    """
+    Adds paper texture/grain to the image.
+
+    Simulates different paper types and ages.
+
+    Parameters:
+        texture_type: Type of texture ("random", "fine", "coarse") (default: "random").
+        intensity: Texture intensity, float or (min, max) tuple (default: (0.05, 0.15)).
+    """
+
+    name = "paper_texture"
+    affects_geometry = False
+
+    def _validate_params(self) -> None:
+        intensity = self.params.get("intensity", (0.05, 0.15))
+        if isinstance(intensity, (int, float)):
+            if not (0 < intensity <= 1):
+                raise ValueError("intensity must be between 0 and 1")
+
+    def apply(
+        self,
+        image: np.ndarray,
+        bboxes: np.ndarray | None = None,
+        rng: np.random.Generator | None = None,
+    ) -> AugmentationResult:
+        rng = rng or np.random.default_rng()
+
+        h, w = image.shape[:2]
+        texture_type = self.params.get("texture_type", "random")
+        intensity = self.params.get("intensity", (0.05, 0.15))
+
+        if texture_type == "random":
+            texture_type = rng.choice(["fine", "coarse"])
+
+        if isinstance(intensity, tuple):
+            intensity = rng.uniform(intensity[0], intensity[1])
+
+        # Generate base noise
+        if texture_type == "fine":
+            # Fine grain texture
+            noise = rng.uniform(-1, 1, (h, w)).astype(np.float32)
+            noise = cv2.GaussianBlur(noise, (3, 3), 0)
+        else:
+            # Coarse texture
+            # Generate at lower resolution and upscale
+            small_h, small_w = h // 4, w // 4
+            noise = rng.uniform(-1, 1, (small_h, small_w)).astype(np.float32)
+            noise = cv2.resize(noise, (w, h), interpolation=cv2.INTER_LINEAR)
+            noise = cv2.GaussianBlur(noise, (5, 5), 0)
+
+        # Apply texture
+        output = image.astype(np.float32)
+        noise_3d = noise[:, :, np.newaxis] * intensity * 255
+        output = output + noise_3d
+
+        output = np.clip(output, 0, 255).astype(np.uint8)
+
+        return AugmentationResult(
+            image=output,
+            bboxes=bboxes.copy() if bboxes is not None else None,
+            metadata={"texture_type": texture_type, "intensity": intensity},
+        )
+
+    def get_preview_params(self) -> dict:
+        return {"texture_type": "coarse", "intensity": 0.15}
+
+
+class ScannerArtifacts(BaseAugmentation):
+    """
+    Adds scanner artifacts to the image.
+
+    Simulates scanner imperfections like lines and dust spots.
+
+    Parameters:
+        line_probability: Probability of adding scan lines (default: 0.3).
+        dust_probability: Probability of adding dust spots (default: 0.4).
+    """
+
+    name = "scanner_artifacts"
+    affects_geometry = False
+
+    def _validate_params(self) -> None:
+        line_prob = self.params.get("line_probability", 0.3)
+        dust_prob = self.params.get("dust_probability", 0.4)
+        if not (0 <= line_prob <= 1):
+            raise ValueError("line_probability must be between 0 and 1")
+        if not (0 <= dust_prob <= 1):
+            raise ValueError("dust_probability must be between 0 and 1")
+
+    def apply(
+        self,
+        image: np.ndarray,
+        bboxes: np.ndarray | None = None,
+        rng: np.random.Generator | None = None,
+    ) -> AugmentationResult:
+        rng = rng or np.random.default_rng()
+
+        h, w = image.shape[:2]
+        line_probability = self.params.get("line_probability", 0.3)
+        dust_probability = self.params.get("dust_probability", 0.4)
+
+        output = image.copy()
+
+        # Add scan lines
+        if rng.random() < line_probability:
+            num_lines = rng.integers(1, 4)
+            for _ in range(num_lines):
+                y = rng.integers(0, h)
+                thickness = rng.integers(1, 3)
+                # Light or dark line
+                color = rng.integers(200, 240) if rng.random() > 0.5 else rng.integers(50, 100)
+
+                # Make line partially transparent
+                alpha = rng.uniform(0.3, 0.6)
+                for dy in range(thickness):
+                    if y + dy < h:
+                        output[y + dy, :] = (
+                            output[y + dy, :].astype(np.float32) * (1 - alpha) +
+                            color * alpha
+                        ).astype(np.uint8)
+
+        # Add dust spots
+        if rng.random() < dust_probability:
+            num_dust = rng.integers(5, 20)
+            for _ in range(num_dust):
+                x = rng.integers(0, w)
+                y = rng.integers(0, h)
+                radius = rng.integers(1, 3)
+
+                # Dark dust spot
+                color = rng.integers(50, 120)
+                cv2.circle(output, (x, y), radius, int(color), -1)
+
+        return AugmentationResult(
+            image=output,
+            bboxes=bboxes.copy() if bboxes is not None else None,
+            metadata={
+                "line_probability": line_probability,
+                "dust_probability": dust_probability,
+            },
+        )
+
+    def get_preview_params(self) -> dict:
+        return {"line_probability": 0.8, "dust_probability": 0.8}
diff --git a/packages/shared/shared/training/__init__.py b/packages/shared/shared/training/__init__.py
new file mode 100644
index 0000000..76d3c08
--- /dev/null
+++ b/packages/shared/shared/training/__init__.py
@@ -0,0 +1,5 @@
+"""Shared training utilities."""
+
+from .yolo_trainer import YOLOTrainer, TrainingConfig, TrainingResult
+
+__all__ = ["YOLOTrainer", "TrainingConfig", "TrainingResult"]
diff --git a/packages/shared/shared/training/yolo_trainer.py b/packages/shared/shared/training/yolo_trainer.py
new file mode 100644
index 0000000..59435cd
--- /dev/null
+++ b/packages/shared/shared/training/yolo_trainer.py
@@ -0,0 +1,239 @@
+"""
+Shared YOLO Training Module
+
+Unified training logic for both CLI and Web API.
+"""
+
+import logging
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import Any, Callable
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class TrainingConfig:
+    """Training configuration."""
+
+    # Model settings
+    model_path: str = "yolo11n.pt"  # Base model or path to trained model
+    data_yaml: str = ""  # Path to data.yaml
+
+    # Training hyperparameters
+    epochs: int = 100
+    batch_size: int = 16
+    image_size: int = 640
+    learning_rate: float = 0.01
+    device: str = "0"
+
+    # Output settings
+    project: str = "runs/train"
+    name: str = "invoice_fields"
+
+    # Performance settings
+    workers: int = 4
+    cache: bool = False
+
+    # Resume settings
+    resume: bool = False
+    resume_from: str | None = None  # Path to checkpoint
+
+    # Document-specific augmentation (optimized for invoices)
+    augmentation: dict[str, Any] = field(default_factory=lambda: {
+        "degrees": 5.0,
+        "translate": 0.05,
+        "scale": 0.2,
+        "shear": 0.0,
+        "perspective": 0.0,
+        "flipud": 0.0,
+        "fliplr": 0.0,
+        "mosaic": 0.0,
+        "mixup": 0.0,
+        "hsv_h": 0.0,
+        "hsv_s": 0.1,
+        "hsv_v": 0.2,
+    })
+
+
+@dataclass
+class TrainingResult:
+    """Training result."""
+
+    success: bool
+    model_path: str | None = None
+    metrics: dict[str, float] = field(default_factory=dict)
+    error: str | None = None
+    save_dir: str | None = None
+
+
+class YOLOTrainer:
+    """Unified YOLO trainer for CLI and Web API."""
+
+    def __init__(
+        self,
+        config: TrainingConfig,
+        log_callback: Callable[[str, str], None] | None = None,
+    ):
+        """
+        Initialize trainer.
+
+        Args:
+            config: Training configuration
+            log_callback: Optional callback for logging (level, message)
+        """
+        self.config = config
+        self._log_callback = log_callback
+
+    def _log(self, level: str, message: str) -> None:
+        """Log a message."""
+        if self._log_callback:
+            self._log_callback(level, message)
+        if level == "INFO":
+            logger.info(message)
+        elif level == "ERROR":
+            logger.error(message)
+        elif level == "WARNING":
+            logger.warning(message)
+
+    def validate_config(self) -> tuple[bool, str | None]:
+        """
+        Validate training configuration.
+
+        Returns:
+            Tuple of (is_valid, error_message)
+        """
+        # Check model path
+        model_path = Path(self.config.model_path)
+        if not model_path.suffix == ".pt":
+            # Could be a model name like "yolo11n.pt" which is downloaded
+            if not model_path.name.startswith("yolo"):
+                return False, f"Invalid model: {self.config.model_path}"
+        elif not model_path.exists():
+            return False, f"Model file not found: {self.config.model_path}"
+
+        # Check data.yaml
+        if not self.config.data_yaml:
+            return False, "data_yaml is required"
+        data_yaml = Path(self.config.data_yaml)
+        if not data_yaml.exists():
+            return False, f"data.yaml not found: {self.config.data_yaml}"
+
+        return True, None
+
+    def train(self) -> TrainingResult:
+        """
+        Run YOLO training.
+
+        Returns:
+            TrainingResult with model path and metrics
+        """
+        try:
+            from ultralytics import YOLO
+        except ImportError:
+            return TrainingResult(
+                success=False,
+                error="Ultralytics (YOLO) not installed. Install with: pip install ultralytics",
+            )
+
+        # Validate config
+        is_valid, error = self.validate_config()
+        if not is_valid:
+            return TrainingResult(success=False, error=error)
+
+        self._log("INFO", f"Starting YOLO training")
+        self._log("INFO", f"  Model: {self.config.model_path}")
+        self._log("INFO", f"  Data: {self.config.data_yaml}")
+        self._log("INFO", f"  Epochs: {self.config.epochs}")
+        self._log("INFO", f"  Batch size: {self.config.batch_size}")
+        self._log("INFO", f"  Image size: {self.config.image_size}")
+
+        try:
+            # Load model
+            if self.config.resume and self.config.resume_from:
+                resume_path = Path(self.config.resume_from)
+                if resume_path.exists():
+                    self._log("INFO", f"Resuming from: {resume_path}")
+                    model = YOLO(str(resume_path))
+                else:
+                    model = YOLO(self.config.model_path)
+            else:
+                model = YOLO(self.config.model_path)
+
+            # Build training arguments
+            train_args = {
+                "data": str(Path(self.config.data_yaml).absolute()),
+                "epochs": self.config.epochs,
+                "batch": self.config.batch_size,
+                "imgsz": self.config.image_size,
+                "lr0": self.config.learning_rate,
+                "device": self.config.device,
+                "project": self.config.project,
+                "name": self.config.name,
+                "exist_ok": True,
+                "pretrained": True,
+                "verbose": True,
+                "workers": self.config.workers,
+                "cache": self.config.cache,
+                "resume": self.config.resume and self.config.resume_from is not None,
+            }
+
+            # Add augmentation settings
+            train_args.update(self.config.augmentation)
+
+            # Train
+            results = model.train(**train_args)
+
+            # Get best model path
+            best_model = Path(results.save_dir) / "weights" / "best.pt"
+
+            # Extract metrics
+            metrics = {}
+            if hasattr(results, "results_dict"):
+                metrics = {
+                    "mAP50": results.results_dict.get("metrics/mAP50(B)", 0),
+                    "mAP50-95": results.results_dict.get("metrics/mAP50-95(B)", 0),
+                    "precision": results.results_dict.get("metrics/precision(B)", 0),
+                    "recall": results.results_dict.get("metrics/recall(B)", 0),
+                }
+
+            self._log("INFO", f"Training completed successfully")
+            self._log("INFO", f"  Best model: {best_model}")
+            self._log("INFO", f"  mAP@0.5: {metrics.get('mAP50', 'N/A')}")
+
+            return TrainingResult(
+                success=True,
+                model_path=str(best_model) if best_model.exists() else None,
+                metrics=metrics,
+                save_dir=str(results.save_dir),
+            )
+
+        except Exception as e:
+            self._log("ERROR", f"Training failed: {e}")
+            return TrainingResult(success=False, error=str(e))
+
+    def validate(self, split: str = "val") -> dict[str, float]:
+        """
+        Run validation on trained model.
+
+        Args:
+            split: Dataset split to validate on ("val" or "test")
+
+        Returns:
+            Validation metrics
+        """
+        try:
+            from ultralytics import YOLO
+
+            model = YOLO(self.config.model_path)
+            metrics = model.val(data=self.config.data_yaml, split=split)
+
+            return {
+                "mAP50": metrics.box.map50,
+                "mAP50-95": metrics.box.map,
+                "precision": metrics.box.mp,
+                "recall": metrics.box.mr,
+            }
+        except Exception as e:
+            self._log("ERROR", f"Validation failed: {e}")
+            return {}
diff --git a/packages/training/training/cli/train.py b/packages/training/training/cli/train.py
index ca64863..fc96997 100644
--- a/packages/training/training/cli/train.py
+++ b/packages/training/training/cli/train.py
@@ -199,67 +199,63 @@ def main():
         db.close()
         return
 
-    # Start training
+    # Start training using shared trainer
     print("\n" + "=" * 60)
     print("Starting YOLO Training")
     print("=" * 60)
 
-    from ultralytics import YOLO
+    from shared.training import YOLOTrainer, TrainingConfig
 
-    # Load model
+    # Determine resume checkpoint
     last_checkpoint = Path(args.project) / args.name / 'weights' / 'last.pt'
-    if args.resume and last_checkpoint.exists():
-        print(f"Resuming from: {last_checkpoint}")
-        model = YOLO(str(last_checkpoint))
-    else:
-        model = YOLO(args.model)
+    resume_from = str(last_checkpoint) if args.resume and last_checkpoint.exists() else None
 
-    # Training arguments
+    # Create training config
     data_yaml = dataset_dir / 'dataset.yaml'
-    train_args = {
-        'data': str(data_yaml.absolute()),
-        'epochs': args.epochs,
-        'batch': args.batch,
-        'imgsz': args.imgsz,
-        'project': args.project,
-        'name': args.name,
-        'device': args.device,
-        'exist_ok': True,
-        'pretrained': True,
-        'verbose': True,
-        'workers': args.workers,
-        'cache': args.cache,
-        'resume': args.resume and last_checkpoint.exists(),
-        # Document-specific augmentation settings
-        'degrees': 5.0,
-        'translate': 0.05,
-        'scale': 0.2,
-        'shear': 0.0,
-        'perspective': 0.0,
-        'flipud': 0.0,
-        'fliplr': 0.0,
-        'mosaic': 0.0,
-        'mixup': 0.0,
-        'hsv_h': 0.0,
-        'hsv_s': 0.1,
-        'hsv_v': 0.2,
-    }
+    config = TrainingConfig(
+        model_path=args.model,
+        data_yaml=str(data_yaml),
+        epochs=args.epochs,
+        batch_size=args.batch,
+        image_size=args.imgsz,
+        device=args.device,
+        project=args.project,
+        name=args.name,
+        workers=args.workers,
+        cache=args.cache,
+        resume=args.resume,
+        resume_from=resume_from,
+    )
 
-    # Train
-    results = model.train(**train_args)
+    # Run training
+    trainer = YOLOTrainer(config=config)
+    result = trainer.train()
+
+    if not result.success:
+        print(f"\nError: Training failed - {result.error}")
+        db.close()
+        sys.exit(1)
 
     # Print results
     print("\n" + "=" * 60)
     print("Training Complete")
     print("=" * 60)
-    print(f"Best model: {args.project}/{args.name}/weights/best.pt")
-    print(f"Last model: {args.project}/{args.name}/weights/last.pt")
+    print(f"Best model: {result.model_path}")
+    print(f"Save directory: {result.save_dir}")
+    if result.metrics:
+        print(f"mAP@0.5: {result.metrics.get('mAP50', 'N/A')}")
+        print(f"mAP@0.5-0.95: {result.metrics.get('mAP50-95', 'N/A')}")
 
     # Validate on test set
     print("\nRunning validation on test set...")
-    metrics = model.val(split='test')
-    print(f"mAP50: {metrics.box.map50:.4f}")
-    print(f"mAP50-95: {metrics.box.map:.4f}")
+    if result.model_path:
+        config.model_path = result.model_path
+        config.data_yaml = str(data_yaml)
+        test_trainer = YOLOTrainer(config=config)
+        test_metrics = test_trainer.validate(split='test')
+        if test_metrics:
+            print(f"mAP50: {test_metrics.get('mAP50', 0):.4f}")
+            print(f"mAP50-95: {test_metrics.get('mAP50-95', 0):.4f}")
 
     # Close database
     db.close()
diff --git a/runs_backup/train/invoice_fields/args.yaml b/runs_backup/train/invoice_fields/args.yaml
new file mode 100644
index 0000000..6aa033a
--- /dev/null
+++ b/runs_backup/train/invoice_fields/args.yaml
@@ -0,0 +1,106 @@
+task: detect
+mode: train
+model: runs/train/invoice_fields/weights/last.pt
+data: /home/kai/invoice-data/dataset/dataset.yaml
+epochs: 100
+time: null
+patience: 100
+batch: 8
+imgsz: 1280
+save: true
+save_period: -1
+cache: false
+device: '0'
+workers: 8
+project: runs/train
+name: invoice_fields
+exist_ok: true
+pretrained: true
+optimizer: auto
+verbose: true
+seed: 0
+deterministic: true
+single_cls: false
+rect: false
+cos_lr: false
+close_mosaic: 10
+resume: runs/train/invoice_fields/weights/last.pt
+amp: true
+fraction: 1.0
+profile: false
+freeze: null
+multi_scale: false
+compile: false
+overlap_mask: true
+mask_ratio: 4
+dropout: 0.0
+val: true
+split: val
+save_json: false
+conf: null
+iou: 0.7
+max_det: 300
+half: false
+dnn: false
+plots: true
+source: null
+vid_stride: 1
+stream_buffer: false
+visualize: false
+augment: false
+agnostic_nms: false
+classes: null
+retina_masks: false
+embed: null
+show: false
+save_frames: false
+save_txt: false
+save_conf: false
+save_crop: false
+show_labels: true
+show_conf: true
+show_boxes: true
+line_width: null
+format: torchscript
+keras: false
+optimize: false
+int8: false
+dynamic: false
+simplify: true
+opset: null
+workspace: null
+nms: false
+lr0: 0.01
+lrf: 0.01
+momentum: 0.937
+weight_decay: 0.0005
+warmup_epochs: 3.0
+warmup_momentum: 0.8
+warmup_bias_lr: 0.0
+box: 7.5
+cls: 0.5
+dfl: 1.5
+pose: 12.0
+kobj: 1.0
+nbs: 64
+hsv_h: 0.0
+hsv_s: 0.1
+hsv_v: 0.2
+degrees: 5.0
+translate: 0.05
+scale: 0.2
+shear: 0.0
+perspective: 0.0
+flipud: 0.0
+fliplr: 0.0
+bgr: 0.0
+mosaic: 0.0
+mixup: 0.0
+cutmix: 0.0
+copy_paste: 0.0
+copy_paste_mode: flip
+auto_augment: randaugment
+erasing: 0.4
+cfg: null
+tracker: botsort.yaml
+save_dir: /mnt/c/Users/yaoji/git/ColaCoder/invoice-master-poc-v2/runs/train/invoice_fields
diff --git a/runs_backup/train/invoice_fields/results.csv b/runs_backup/train/invoice_fields/results.csv
new file mode 100644
index 0000000..2a35a32
--- /dev/null
+++ b/runs_backup/train/invoice_fields/results.csv
@@ -0,0 +1,101 @@
+epoch,time,train/box_loss,train/cls_loss,train/dfl_loss,metrics/precision(B),metrics/recall(B),metrics/mAP50(B),metrics/mAP50-95(B),val/box_loss,val/cls_loss,val/dfl_loss,lr/pg0,lr/pg1,lr/pg2
+1,507.213,0.97459,2.44056,1.09407,0.62221,0.6847,0.6581,0.53505,0.66652,1.13762,0.89651,0.00333127,0.00333127,0.00333127
+2,960.144,0.65449,1.06648,0.90865,0.62998,0.73947,0.70066,0.58144,0.63394,1.03248,0.89183,0.00659863,0.00659863,0.00659863
+3,1415.64,0.64921,0.97725,0.90901,0.70271,0.7713,0.76974,0.60362,0.7453,0.92005,0.9044,0.00979998,0.00979998,0.00979998
+4,1871.6,0.61159,0.90785,0.90192,0.7264,0.77507,0.78958,0.66909,0.57561,0.83662,0.87737,0.009703,0.009703,0.009703
+5,2317.68,0.55503,0.81107,0.88628,0.75742,0.81685,0.8287,0.70772,0.57193,0.74866,0.87531,0.009604,0.009604,0.009604
+6,2756.85,0.52443,0.75067,0.8767,0.77362,0.82743,0.84758,0.73413,0.53604,0.70643,0.86947,0.009505,0.009505,0.009505
+7,3211.38,0.49889,0.70526,0.8696,0.79041,0.83537,0.86367,0.75099,0.52784,0.67021,0.86686,0.009406,0.009406,0.009406
+8,3657.41,0.47889,0.66715,0.86261,0.80402,0.85067,0.87834,0.7573,0.5365,0.64019,0.86754,0.009307,0.009307,0.009307
+9,4100.61,0.46417,0.63485,0.85761,0.81776,0.84993,0.88612,0.77638,0.50423,0.61379,0.85539,0.009208,0.009208,0.009208
+10,4544.47,0.45154,0.61098,0.85334,0.82044,0.86313,0.89168,0.7685,0.54578,0.59853,0.8668,0.009109,0.009109,0.009109
+11,4983.87,0.44,0.58833,0.84961,0.82995,0.87408,0.90194,0.7972,0.48628,0.57037,0.84686,0.00901,0.00901,0.00901
+12,5419.2,0.42826,0.57051,0.84622,0.83627,0.87342,0.90482,0.79702,0.49356,0.55869,0.84836,0.008911,0.008911,0.008911
+13,5854.78,0.42052,0.55192,0.84327,0.83405,0.88243,0.90883,0.79807,0.50137,0.54503,0.85114,0.008812,0.008812,0.008812
+14,6288.44,0.41497,0.53873,0.84105,0.83687,0.8769,0.90767,0.79881,0.49873,0.54655,0.85078,0.008713,0.008713,0.008713
+15,6724.06,0.40759,0.52501,0.83904,0.84211,0.88472,0.91329,0.81111,0.47871,0.52982,0.84461,0.008614,0.008614,0.008614
+16,7164.64,0.39956,0.51303,0.83591,0.85078,0.88609,0.91766,0.80948,0.49199,0.52052,0.84634,0.008515,0.008515,0.008515
+17,7626.93,0.39651,0.50255,0.83453,0.84894,0.89151,0.92036,0.80185,0.5227,0.52229,0.8514,0.008416,0.008416,0.008416
+18,490.978,0.3953,0.49497,0.83435,0.86113,0.87974,0.92114,0.80521,0.51284,0.51491,0.84926,0.008317,0.008317,0.008317
+19,976.368,0.39234,0.49242,0.8328,0.86521,0.88431,0.92347,0.80898,0.5037,0.50744,0.84824,0.008218,0.008218,0.008218
+20,1457.68,0.3864,0.48591,0.8308,0.86361,0.88895,0.92337,0.80562,0.51844,0.50692,0.8497,0.008119,0.008119,0.008119
+21,1938.95,0.3826,0.47627,0.83086,0.85282,0.89488,0.92276,0.80107,0.53348,0.50706,0.85418,0.00802,0.00802,0.00802
+22,2418,0.37996,0.46715,0.82911,0.86698,0.89705,0.92757,0.81493,0.5038,0.49105,0.84342,0.007921,0.007921,0.007921
+23,2900.33,0.37555,0.46089,0.82769,0.86871,0.89566,0.92949,0.82251,0.48233,0.48473,0.84078,0.007822,0.007822,0.007822
+24,3381.33,0.3717,0.45531,0.82676,0.87203,0.89509,0.93002,0.80968,0.52892,0.49231,0.85017,0.007723,0.007723,0.007723
+25,3867.95,0.36913,0.44591,0.82544,0.87636,0.89074,0.93018,0.81083,0.5286,0.4882,0.84889,0.007624,0.007624,0.007624
+26,4354.94,0.3662,0.44016,0.82483,0.86957,0.89807,0.92963,0.80554,0.5486,0.49546,0.8527,0.007525,0.007525,0.007525
+27,4839.34,0.36364,0.43794,0.82368,0.87515,0.89602,0.93102,0.80927,0.54175,0.49001,0.85159,0.007426,0.007426,0.007426
+28,5324.36,0.36043,0.42951,0.8234,0.87074,0.90178,0.93175,0.8107,0.53742,0.4862,0.85096,0.007327,0.007327,0.007327
+29,5810.68,0.35852,0.42834,0.82242,0.87574,0.8992,0.93108,0.80885,0.53965,0.48712,0.85055,0.007228,0.007228,0.007228
+30,6294.55,0.35572,0.42015,0.82155,0.87635,0.90084,0.93172,0.81312,0.5252,0.48089,0.84711,0.007129,0.007129,0.007129
+31,6778.25,0.3539,0.41622,0.82025,0.8777,0.90106,0.93275,0.81359,0.52726,0.47864,0.84712,0.00703,0.00703,0.00703
+32,7261.88,0.35095,0.4105,0.82048,0.87684,0.90364,0.9337,0.81598,0.52269,0.47547,0.84625,0.006931,0.006931,0.006931
+33,7745.73,0.34835,0.40539,0.81904,0.87629,0.90454,0.933,0.81501,0.52381,0.47649,0.84713,0.006832,0.006832,0.006832
+34,8234.78,0.34699,0.402,0.8182,0.8755,0.90506,0.93336,0.81532,0.52435,0.4756,0.84749,0.006733,0.006733,0.006733
+35,8716.45,0.34487,0.39711,0.81704,0.87551,0.90432,0.93325,0.81529,0.52489,0.47558,0.84763,0.006634,0.006634,0.006634
+36,9200.09,0.34343,0.39638,0.81749,0.87551,0.90495,0.93343,0.81477,0.52659,0.4765,0.84798,0.006535,0.006535,0.006535
+37,9683.76,0.34013,0.3907,0.81647,0.87738,0.90416,0.93371,0.8152,0.52755,0.47568,0.84807,0.006436,0.006436,0.006436
+38,10164.4,0.33826,0.38626,0.81468,0.87923,0.90342,0.93399,0.81571,0.52756,0.47527,0.84794,0.006337,0.006337,0.006337
+39,10648.6,0.3366,0.3812,0.81517,0.8786,0.90333,0.93368,0.81448,0.53035,0.47601,0.84857,0.006238,0.006238,0.006238
+40,11130.6,0.3353,0.37879,0.8151,0.87974,0.90405,0.93411,0.81544,0.52739,0.47333,0.84768,0.006139,0.006139,0.006139
+41,11612.6,0.33395,0.37397,0.8143,0.88034,0.90315,0.93432,0.81514,0.529,0.47304,0.84765,0.00604,0.00604,0.00604
+42,12097.5,0.33104,0.37164,0.81448,0.87942,0.90449,0.93429,0.81484,0.53077,0.47386,0.84799,0.005941,0.005941,0.005941
+43,12579.8,0.33016,0.3681,0.81305,0.87964,0.90412,0.93457,0.81473,0.53129,0.47382,0.84789,0.005842,0.005842,0.005842
+44,13065.3,0.32845,0.36431,0.81191,0.88092,0.90312,0.9348,0.81538,0.53025,0.47337,0.84751,0.005743,0.005743,0.005743
+45,13550.4,0.32642,0.36127,0.81295,0.8805,0.905,0.93502,0.81544,0.53124,0.47333,0.8477,0.005644,0.005644,0.005644
+46,14034.3,0.32483,0.35761,0.81158,0.87898,0.90636,0.935,0.81556,0.53135,0.47317,0.84772,0.005545,0.005545,0.005545
+47,14517.3,0.32236,0.35337,0.81014,0.88018,0.90502,0.93493,0.81547,0.53228,0.473,0.8478,0.005446,0.005446,0.005446
+48,14998.6,0.3211,0.35051,0.81064,0.87941,0.9055,0.93481,0.81473,0.5353,0.47335,0.84839,0.005347,0.005347,0.005347
+49,15479.2,0.32043,0.34797,0.8097,0.87884,0.90584,0.93482,0.81429,0.53741,0.47359,0.84867,0.005248,0.005248,0.005248
+50,15962,0.3182,0.34589,0.80867,0.87776,0.90777,0.93476,0.81395,0.53841,0.47358,0.84884,0.005149,0.005149,0.005149
+51,16445.4,0.31722,0.34332,0.80879,0.87932,0.90605,0.93488,0.81439,0.53827,0.47301,0.84874,0.00505,0.00505,0.00505
+52,16925.3,0.31437,0.33963,0.80846,0.87925,0.90688,0.93521,0.81468,0.53772,0.47254,0.84861,0.004951,0.004951,0.004951
+53,17410.4,0.31435,0.33632,0.80792,0.88015,0.9069,0.93538,0.8148,0.53789,0.47193,0.84865,0.004852,0.004852,0.004852
+54,17893.2,0.31341,0.33352,0.80833,0.88132,0.90605,0.93552,0.81465,0.53934,0.47218,0.84889,0.004753,0.004753,0.004753
+55,18375.2,0.31091,0.33239,0.80688,0.88151,0.90617,0.93547,0.81453,0.54019,0.47227,0.84885,0.004654,0.004654,0.004654
+56,18857.4,0.30906,0.32753,0.80616,0.88348,0.90504,0.93565,0.81495,0.53974,0.47179,0.84864,0.004555,0.004555,0.004555
+57,19340.6,0.30722,0.32334,0.80582,0.88502,0.90326,0.93558,0.81484,0.53977,0.47174,0.84859,0.004456,0.004456,0.004456
+58,19824.8,0.30575,0.3195,0.80592,0.88348,0.90487,0.93566,0.81511,0.53921,0.47158,0.84832,0.004357,0.004357,0.004357
+59,20305.7,0.30426,0.31846,0.80494,0.88419,0.90477,0.93575,0.81534,0.5387,0.47138,0.84819,0.004258,0.004258,0.004258
+60,20785.1,0.30295,0.3154,0.80494,0.88302,0.90624,0.93568,0.81572,0.53769,0.47106,0.84788,0.004159,0.004159,0.004159
+61,21263.8,0.3013,0.3131,0.80436,0.88438,0.90545,0.93572,0.81606,0.53622,0.47079,0.84762,0.00406,0.00406,0.00406
+62,21746,0.30019,0.31077,0.80391,0.88296,0.90732,0.93578,0.8165,0.53455,0.47011,0.84724,0.003961,0.003961,0.003961
+63,22225,0.29841,0.30656,0.80379,0.88244,0.90779,0.93591,0.81693,0.53417,0.47003,0.84715,0.003862,0.003862,0.003862
+64,22704.7,0.29696,0.30489,0.80284,0.88493,0.90596,0.9359,0.81716,0.53362,0.47013,0.84707,0.003763,0.003763,0.003763
+65,23182.9,0.2952,0.30022,0.80288,0.88366,0.90731,0.93594,0.81737,0.533,0.47024,0.84695,0.003664,0.003664,0.003664
+66,23663.6,0.29337,0.29898,0.80273,0.88514,0.90611,0.93609,0.81805,0.53189,0.47015,0.84664,0.003565,0.003565,0.003565
+67,24149.7,0.29248,0.29492,0.80242,0.88664,0.90536,0.93607,0.81783,0.53208,0.4704,0.84665,0.003466,0.003466,0.003466
+68,24629,0.28987,0.29155,0.8009,0.89014,0.90207,0.93611,0.81811,0.53203,0.47046,0.84656,0.003367,0.003367,0.003367
+69,25109.9,0.28942,0.29004,0.8011,0.88939,0.90353,0.93619,0.81872,0.53127,0.4704,0.84631,0.003268,0.003268,0.003268
+70,25590.5,0.28752,0.28571,0.80059,0.88926,0.90393,0.93627,0.81909,0.53074,0.47023,0.84624,0.003169,0.003169,0.003169
+71,26072.8,0.28546,0.28301,0.7999,0.88844,0.90476,0.93631,0.81967,0.52999,0.47005,0.84612,0.00307,0.00307,0.00307
+72,26552.9,0.2842,0.28027,0.79942,0.88801,0.90505,0.93622,0.81978,0.52939,0.46994,0.84607,0.002971,0.002971,0.002971
+73,27035.5,0.28297,0.27907,0.79956,0.88781,0.90499,0.93615,0.82032,0.528,0.4694,0.84578,0.002872,0.002872,0.002872
+74,27518.8,0.2812,0.27446,0.79886,0.88848,0.90493,0.93611,0.82061,0.52675,0.46906,0.84549,0.002773,0.002773,0.002773
+75,28007.4,0.27866,0.27202,0.79684,0.88889,0.90467,0.9361,0.82099,0.5257,0.4692,0.84529,0.002674,0.002674,0.002674
+76,28499.1,0.27708,0.26798,0.7978,0.88807,0.9054,0.93615,0.82138,0.52574,0.46928,0.84523,0.002575,0.002575,0.002575
+77,28993.2,0.27398,0.2644,0.79612,0.88825,0.9055,0.93616,0.82161,0.52491,0.46925,0.84496,0.002476,0.002476,0.002476
+78,29480.5,0.27359,0.26209,0.79678,0.88876,0.90547,0.93617,0.82172,0.52467,0.4691,0.84498,0.002377,0.002377,0.002377
+79,29970.7,0.27153,0.25905,0.79585,0.88942,0.90548,0.93613,0.82211,0.52407,0.46892,0.84474,0.002278,0.002278,0.002278
+80,30453.8,0.2696,0.25647,0.79513,0.88897,0.90665,0.93617,0.82246,0.52298,0.46881,0.84445,0.002179,0.002179,0.002179
+81,30936.1,0.26895,0.25375,0.79466,0.88799,0.90791,0.93617,0.8226,0.52253,0.46904,0.84445,0.00208,0.00208,0.00208
+82,31418.6,0.26733,0.25025,0.79474,0.88945,0.90695,0.93608,0.82293,0.52172,0.4694,0.84434,0.001981,0.001981,0.001981
+83,31911.4,0.26537,0.24754,0.79479,0.89112,0.90496,0.93604,0.82338,0.52094,0.46932,0.84421,0.001882,0.001882,0.001882
+84,32402,0.26344,0.24514,0.79369,0.8928,0.90319,0.93598,0.82353,0.52015,0.46957,0.84407,0.001783,0.001783,0.001783
+85,32903.1,0.26045,0.24052,0.79226,0.89211,0.90347,0.93615,0.82427,0.51861,0.46958,0.84372,0.001684,0.001684,0.001684
+86,33413.6,0.25867,0.23781,0.79209,0.89286,0.90279,0.9362,0.82493,0.51664,0.47018,0.84338,0.001585,0.001585,0.001585
+87,33923.6,0.257,0.23463,0.792,0.89299,0.90297,0.93614,0.8254,0.5147,0.46974,0.84305,0.001486,0.001486,0.001486
+88,34436.4,0.25569,0.23153,0.79149,0.89278,0.90277,0.93609,0.82622,0.51242,0.4697,0.84266,0.001387,0.001387,0.001387
+89,34949.6,0.25343,0.22868,0.791,0.89137,0.90434,0.93599,0.82675,0.51036,0.46947,0.84227,0.001288,0.001288,0.001288
+90,35449.6,0.25194,0.22502,0.79051,0.89128,0.90489,0.93591,0.82729,0.5092,0.46975,0.84219,0.001189,0.001189,0.001189
+91,35960.5,0.2502,0.22225,0.78959,0.8898,0.90646,0.93586,0.82781,0.50761,0.46999,0.84186,0.00109,0.00109,0.00109
+92,36452.1,0.24777,0.21844,0.78906,0.89054,0.9057,0.93593,0.82831,0.50603,0.47043,0.84154,0.000991,0.000991,0.000991
+93,36942.9,0.24554,0.21503,0.78861,0.88979,0.90679,0.93584,0.82858,0.50495,0.4703,0.84125,0.000892,0.000892,0.000892
+94,37430.3,0.2434,0.21193,0.78799,0.88928,0.90756,0.93566,0.8288,0.5041,0.47075,0.84109,0.000793,0.000793,0.000793
+95,37918.7,0.2413,0.20892,0.78736,0.8899,0.90683,0.93567,0.82882,0.50339,0.47152,0.84105,0.000694,0.000694,0.000694
+96,38404.3,0.2405,0.20619,0.78595,0.88999,0.90713,0.9355,0.82912,0.50244,0.47239,0.84104,0.000595,0.000595,0.000595
+97,38893.8,0.23808,0.2031,0.78683,0.88982,0.90634,0.93531,0.82938,0.50187,0.47281,0.84116,0.000496,0.000496,0.000496
+98,39382.8,0.23581,0.20034,0.78643,0.89144,0.9045,0.93517,0.82959,0.50119,0.47383,0.84128,0.000397,0.000397,0.000397
+99,39871.7,0.23432,0.19778,0.78568,0.89187,0.90415,0.93488,0.82953,0.50058,0.47452,0.84126,0.000298,0.000298,0.000298
+100,40359.7,0.233,0.19485,0.78528,0.89228,0.90349,0.93471,0.82961,0.50029,0.47497,0.84139,0.000199,0.000199,0.000199
diff --git a/runs_backup/train/invoice_yolo11n_full/args.yaml b/runs_backup/train/invoice_yolo11n_full/args.yaml
new file mode 100644
index 0000000..19a11c5
--- /dev/null
+++ b/runs_backup/train/invoice_yolo11n_full/args.yaml
@@ -0,0 +1,106 @@
+task: detect
+mode: train
+model: yolo11n.pt
+data: /home/kai/invoice-data/dataset/dataset.yaml
+epochs: 100
+time: null
+patience: 100
+batch: 16
+imgsz: 1280
+save: true
+save_period: -1
+cache: false
+device: '0'
+workers: 8
+project: runs/train
+name: invoice_yolo11n_full
+exist_ok: true
+pretrained: true
+optimizer: auto
+verbose: true
+seed: 0
+deterministic: true
+single_cls: false
+rect: false
+cos_lr: false
+close_mosaic: 10
+resume: false
+amp: true
+fraction: 1.0
+profile: false
+freeze: null
+multi_scale: false
+compile: false
+overlap_mask: true
+mask_ratio: 4
+dropout: 0.0
+val: true
+split: val
+save_json: false
+conf: null
+iou: 0.7
+max_det: 300
+half: false
+dnn: false
+plots: true
+source: null
+vid_stride: 1
+stream_buffer: false
+visualize: false
+augment: false
+agnostic_nms: false
+classes: null
+retina_masks: false
+embed: null
+show: false
+save_frames: false
+save_txt: false
+save_conf: false
+save_crop: false
+show_labels: true
+show_conf: true
+show_boxes: true
+line_width: null
+format: torchscript
+keras: false
+optimize: false
+int8: false
+dynamic: false
+simplify: true
+opset: null
+workspace: null
+nms: false
+lr0: 0.01
+lrf: 0.01
+momentum: 0.937
+weight_decay: 0.0005
+warmup_epochs: 3.0
+warmup_momentum: 0.8
+warmup_bias_lr: 0.1
+box: 7.5
+cls: 0.5
+dfl: 1.5
+pose: 12.0
+kobj: 1.0
+nbs: 64
+hsv_h: 0.0
+hsv_s: 0.1
+hsv_v: 0.2
+degrees: 5.0
+translate: 0.05
+scale: 0.2
+shear: 0.0
+perspective: 0.0
+flipud: 0.0
+fliplr: 0.0
+bgr: 0.0
+mosaic: 0.0
+mixup: 0.0
+cutmix: 0.0
+copy_paste: 0.0
+copy_paste_mode: flip
+auto_augment: randaugment
+erasing: 0.4
+cfg: null
+tracker: botsort.yaml
+save_dir: /mnt/c/Users/yaoji/git/ColaCoder/invoice-master-poc-v2/runs/train/invoice_yolo11n_full
diff --git a/runs_backup/train/invoice_yolo11n_full/results.csv b/runs_backup/train/invoice_yolo11n_full/results.csv
new file mode 100644
index 0000000..181dab6
--- /dev/null
+++ b/runs_backup/train/invoice_yolo11n_full/results.csv
@@ -0,0 +1,101 @@
+epoch,time,train/box_loss,train/cls_loss,train/dfl_loss,metrics/precision(B),metrics/recall(B),metrics/mAP50(B),metrics/mAP50-95(B),val/box_loss,val/cls_loss,val/dfl_loss,lr/pg0,lr/pg1,lr/pg2
+1,217.641,0.79856,2.56507,1.01986,0.8921,0.84545,0.90033,0.81508,0.49347,0.92369,0.83815,0.00332991,0.00332991,0.00332991
+2,410.275,0.506,1.04596,0.85661,0.9164,0.85418,0.92852,0.79726,0.63851,0.72277,0.87404,0.00659728,0.00659728,0.00659728
+3,598.713,0.49618,0.68647,0.85624,0.93775,0.80014,0.91481,0.80383,0.5798,0.76691,0.86835,0.00979865,0.00979865,0.00979865
+4,782.868,0.44059,0.53532,0.84299,0.94668,0.90421,0.96101,0.8832,0.43961,0.49649,0.84298,0.009703,0.009703,0.009703
+5,967.898,0.37596,0.44308,0.82667,0.88316,0.81492,0.91376,0.82272,0.50616,0.70202,0.85825,0.009604,0.009604,0.009604
+6,1152.03,0.33999,0.39482,0.81661,0.81567,0.73691,0.82644,0.75085,0.43451,0.92038,0.86158,0.009505,0.009505,0.009505
+7,1335.35,0.31114,0.35971,0.80992,0.95256,0.89807,0.96383,0.84241,0.60248,0.48455,0.88156,0.009406,0.009406,0.009406
+8,1518.68,0.29176,0.33987,0.80516,0.97058,0.91185,0.97221,0.85356,0.58408,0.43771,0.86239,0.009307,0.009307,0.009307
+9,1702.03,0.27683,0.3214,0.80166,0.96403,0.91663,0.9736,0.85118,0.6359,0.43055,0.88091,0.009208,0.009208,0.009208
+10,1891.76,0.26487,0.30796,0.79943,0.96201,0.92669,0.97715,0.894,0.46381,0.37437,0.84314,0.009109,0.009109,0.009109
+11,2081.79,0.25744,0.29846,0.79614,0.96562,0.92382,0.97554,0.79415,0.81302,0.46321,0.93576,0.00901,0.00901,0.00901
+12,2273.34,0.24726,0.28842,0.79445,0.96248,0.92901,0.97544,0.7642,0.93193,0.48807,0.98769,0.008911,0.008911,0.008911
+13,2461.71,0.24266,0.27619,0.79413,0.9672,0.93016,0.97834,0.69208,1.24698,0.59927,1.19042,0.008812,0.008812,0.008812
+14,2649.81,0.23391,0.26941,0.79165,0.96579,0.93247,0.98028,0.72182,1.0815,0.50846,1.08505,0.008713,0.008713,0.008713
+15,2837.95,0.22893,0.2651,0.79082,0.9639,0.93414,0.9807,0.8522,0.64306,0.38931,0.88909,0.008614,0.008614,0.008614
+16,3021.27,0.22269,0.25369,0.78809,0.97667,0.92754,0.98283,0.79198,0.89233,0.43512,0.98623,0.008515,0.008515,0.008515
+17,3209.3,0.21937,0.24886,0.78797,0.96559,0.93193,0.98178,0.67198,1.35518,0.59949,1.30325,0.008416,0.008416,0.008416
+18,3400.76,0.21415,0.24489,0.78789,0.95973,0.94489,0.98156,0.63227,1.45967,0.63486,1.42858,0.008317,0.008317,0.008317
+19,3590.96,0.21227,0.23986,0.78736,0.96136,0.94369,0.98263,0.83035,0.76379,0.39831,0.92541,0.008218,0.008218,0.008218
+20,3779.15,0.20834,0.23506,0.78475,0.96214,0.93563,0.97908,0.5024,1.92081,0.81282,1.99717,0.008119,0.008119,0.008119
+21,3976.3,0.20592,0.23055,0.78534,0.9636,0.94141,0.98186,0.71087,1.20783,0.53596,1.17998,0.00802,0.00802,0.00802
+22,4165.69,0.20195,0.22554,0.78431,0.96621,0.94394,0.98458,0.86353,0.6245,0.35194,0.86978,0.007921,0.007921,0.007921
+23,4357.69,0.19847,0.22066,0.78362,0.9745,0.93877,0.98501,0.84365,0.71155,0.37717,0.91395,0.007822,0.007822,0.007822
+24,4548.46,0.19715,0.21991,0.78423,0.96456,0.94907,0.98541,0.77136,0.9901,0.45056,1.02401,0.007723,0.007723,0.007723
+25,4738.2,0.19207,0.21284,0.7821,0.97136,0.94417,0.98568,0.8139,0.83053,0.41526,0.94475,0.007624,0.007624,0.007624
+26,4926.95,0.19124,0.21138,0.7823,0.9712,0.94333,0.98466,0.78106,0.94702,0.44977,1.0068,0.007525,0.007525,0.007525
+27,5115.34,0.18944,0.21166,0.78245,0.97207,0.9347,0.98325,0.57941,1.64865,0.72474,1.70082,0.007426,0.007426,0.007426
+28,5303.92,0.18817,0.20777,0.7814,0.96837,0.9519,0.98672,0.77596,0.96734,0.43218,1.01592,0.007327,0.007327,0.007327
+29,5493.53,0.18565,0.20231,0.78154,0.9719,0.94481,0.98552,0.67875,1.27094,0.57309,1.27411,0.007228,0.007228,0.007228
+30,5682.31,0.18424,0.19916,0.77989,0.96714,0.95269,0.98588,0.73712,1.08764,0.52054,1.09884,0.007129,0.007129,0.007129
+31,5870.11,0.1812,0.19544,0.78013,0.9698,0.95028,0.98687,0.72258,1.15995,0.50918,1.14823,0.00703,0.00703,0.00703
+32,6060.08,0.1801,0.19571,0.7799,0.9699,0.95342,0.98761,0.83372,0.70788,0.36986,0.89674,0.006931,0.006931,0.006931
+33,6244.87,0.17816,0.19373,0.7789,0.96278,0.95546,0.98684,0.85265,0.66817,0.36063,0.88589,0.006832,0.006832,0.006832
+34,6427.13,0.176,0.1909,0.77864,0.96962,0.95053,0.98707,0.85351,0.70717,0.36925,0.90913,0.006733,0.006733,0.006733
+35,6619.57,0.17377,0.18513,0.77794,0.97418,0.94725,0.98662,0.83346,0.77385,0.3987,0.92964,0.006634,0.006634,0.006634
+36,6807.72,0.17359,0.18454,0.77885,0.97363,0.95,0.98703,0.85072,0.72917,0.36745,0.91556,0.006535,0.006535,0.006535
+37,7001.26,0.17126,0.18179,0.77796,0.96337,0.95646,0.98744,0.79259,0.86734,0.4072,0.96666,0.006436,0.006436,0.006436
+38,7186.91,0.16989,0.17967,0.77791,0.97277,0.94891,0.98737,0.8268,0.78213,0.38811,0.9293,0.006337,0.006337,0.006337
+39,7372.72,0.16823,0.17959,0.77698,0.96961,0.95206,0.98714,0.86035,0.69764,0.35696,0.90589,0.006238,0.006238,0.006238
+40,7558.51,0.16639,0.17648,0.77676,0.96471,0.9592,0.98756,0.85427,0.70458,0.35248,0.90042,0.006139,0.006139,0.006139
+41,7747.05,0.16641,0.17472,0.77686,0.96565,0.95698,0.98751,0.76422,1.01303,0.45638,1.04876,0.00604,0.00604,0.00604
+42,7930.66,0.16528,0.17295,0.77783,0.98029,0.94412,0.98709,0.68134,1.27076,0.55229,1.27714,0.005941,0.005941,0.005941
+43,8113.31,0.16304,0.17093,0.77627,0.96909,0.95623,0.98692,0.77479,0.97729,0.45248,1.03101,0.005842,0.005842,0.005842
+44,8298.3,0.16163,0.16817,0.77509,0.96809,0.9575,0.98709,0.80448,0.85637,0.40945,0.96298,0.005743,0.005743,0.005743
+45,8485.25,0.16053,0.16768,0.77535,0.97311,0.95211,0.98726,0.8047,0.85835,0.40888,0.96605,0.005644,0.005644,0.005644
+46,8669.88,0.15959,0.16634,0.77576,0.97431,0.95186,0.98739,0.797,0.87218,0.41446,0.97186,0.005545,0.005545,0.005545
+47,8853.93,0.15778,0.16234,0.77599,0.97532,0.95052,0.98702,0.78582,0.92511,0.43102,1.0039,0.005446,0.005446,0.005446
+48,9037.52,0.15602,0.16175,0.77439,0.97529,0.94998,0.9874,0.84071,0.77064,0.38361,0.93995,0.005347,0.005347,0.005347
+49,9223.9,0.15478,0.1604,0.77364,0.97345,0.95143,0.98729,0.84662,0.73185,0.37143,0.92248,0.005248,0.005248,0.005248
+50,9411.45,0.15431,0.1584,0.77449,0.98033,0.94592,0.98733,0.86995,0.63137,0.34173,0.8816,0.005149,0.005149,0.005149
+51,9595.98,0.15305,0.15648,0.77414,0.97318,0.95431,0.98753,0.86938,0.63298,0.34305,0.88158,0.00505,0.00505,0.00505
+52,9779.61,0.15291,0.15561,0.77441,0.97824,0.94936,0.98777,0.87333,0.60174,0.33831,0.87135,0.004951,0.004951,0.004951
+53,9963.17,0.15193,0.15454,0.77361,0.97077,0.95579,0.98737,0.87038,0.62864,0.34404,0.87819,0.004852,0.004852,0.004852
+54,10151.9,0.14978,0.15218,0.77431,0.97892,0.94812,0.9874,0.86664,0.6559,0.35033,0.88752,0.004753,0.004753,0.004753
+55,10335.2,0.14867,0.14954,0.77318,0.98227,0.94556,0.98734,0.86535,0.66191,0.35233,0.89052,0.004654,0.004654,0.004654
+56,10517.7,0.14781,0.1504,0.77387,0.97187,0.95472,0.98731,0.85393,0.70291,0.36551,0.90638,0.004555,0.004555,0.004555
+57,10704.4,0.14704,0.14654,0.77286,0.96973,0.95539,0.98731,0.84386,0.75517,0.38619,0.92774,0.004456,0.004456,0.004456
+58,10888.6,0.14478,0.14588,0.77324,0.9792,0.94676,0.9872,0.84023,0.76095,0.38846,0.93011,0.004357,0.004357,0.004357
+59,11071.2,0.14408,0.14418,0.7724,0.9709,0.95553,0.98729,0.8499,0.71784,0.37089,0.91332,0.004258,0.004258,0.004258
+60,11256.4,0.1427,0.14165,0.77106,0.96919,0.95682,0.98729,0.85156,0.70256,0.36509,0.90774,0.004159,0.004159,0.004159
+61,11444.8,0.14194,0.14087,0.77269,0.96601,0.96121,0.98731,0.85107,0.70839,0.36753,0.90948,0.00406,0.00406,0.00406
+62,11630.9,0.14062,0.13882,0.77215,0.96628,0.96081,0.98736,0.84858,0.73074,0.3762,0.92033,0.003961,0.003961,0.003961
+63,11816.5,0.13938,0.13865,0.77152,0.96711,0.95961,0.98744,0.85214,0.70862,0.36754,0.91079,0.003862,0.003862,0.003862
+64,12005.3,0.13858,0.13687,0.77045,0.96702,0.9595,0.98748,0.85574,0.69672,0.36084,0.90588,0.003763,0.003763,0.003763
+65,12191.8,0.13775,0.13411,0.77132,0.96785,0.95943,0.9874,0.85729,0.68875,0.35766,0.90221,0.003664,0.003664,0.003664
+66,12379.6,0.13556,0.13271,0.77167,0.96725,0.96005,0.98735,0.85898,0.68174,0.3561,0.89887,0.003565,0.003565,0.003565
+67,12565.4,0.13463,0.13108,0.77009,0.97381,0.95338,0.98732,0.86031,0.67263,0.35399,0.89484,0.003466,0.003466,0.003466
+68,12752.5,0.13515,0.1311,0.77095,0.96906,0.95916,0.98725,0.86029,0.66717,0.35274,0.89292,0.003367,0.003367,0.003367
+69,12940.8,0.13415,0.12957,0.76963,0.97126,0.95685,0.9873,0.86049,0.6644,0.35306,0.8918,0.003268,0.003268,0.003268
+70,13133.2,0.13179,0.12737,0.76937,0.97287,0.95478,0.98727,0.86047,0.6632,0.35246,0.89193,0.003169,0.003169,0.003169
+71,13319.4,0.13185,0.1274,0.77079,0.97267,0.95587,0.98722,0.86193,0.65949,0.35213,0.89086,0.00307,0.00307,0.00307
+72,13504.8,0.12947,0.12446,0.76998,0.97199,0.95686,0.98725,0.86401,0.64895,0.34877,0.88741,0.002971,0.002971,0.002971
+73,13695.1,0.12876,0.12321,0.76883,0.9723,0.9569,0.98725,0.86643,0.64091,0.3473,0.88447,0.002872,0.002872,0.002872
+74,13882,0.12828,0.12194,0.76915,0.97256,0.95686,0.98732,0.86847,0.6322,0.34702,0.88109,0.002773,0.002773,0.002773
+75,14075,0.12664,0.11944,0.76878,0.97277,0.95678,0.98726,0.86861,0.63123,0.3482,0.88086,0.002674,0.002674,0.002674
+76,14259.9,0.12587,0.11965,0.7692,0.9727,0.95673,0.98717,0.86916,0.62721,0.34811,0.87932,0.002575,0.002575,0.002575
+77,14451.2,0.12433,0.1174,0.76838,0.97267,0.95663,0.9872,0.87057,0.62032,0.34709,0.8769,0.002476,0.002476,0.002476
+78,14636.3,0.12352,0.11507,0.76971,0.97087,0.95829,0.98721,0.87189,0.61271,0.34667,0.87445,0.002377,0.002377,0.002377
+79,14821.1,0.1231,0.11454,0.76897,0.97195,0.95752,0.98722,0.87292,0.60714,0.34596,0.87271,0.002278,0.002278,0.002278
+80,15007.3,0.12117,0.11285,0.76864,0.97146,0.95789,0.98726,0.8735,0.6031,0.34515,0.87163,0.002179,0.002179,0.002179
+81,15199.5,0.12029,0.11158,0.76708,0.97018,0.95938,0.9872,0.87378,0.60116,0.34538,0.87113,0.00208,0.00208,0.00208
+82,15390.8,0.11877,0.10949,0.76719,0.97021,0.95964,0.98721,0.87422,0.59897,0.3457,0.87057,0.001981,0.001981,0.001981
+83,15577.6,0.11812,0.10818,0.76749,0.97013,0.95951,0.98722,0.87429,0.59878,0.34524,0.87054,0.001882,0.001882,0.001882
+84,15761.5,0.11687,0.10634,0.76703,0.97155,0.9583,0.98713,0.87407,0.59964,0.34532,0.8709,0.001783,0.001783,0.001783
+85,15946.2,0.11551,0.10455,0.7672,0.9717,0.95797,0.9871,0.87367,0.60049,0.34569,0.87136,0.001684,0.001684,0.001684
+86,16130.6,0.11474,0.10479,0.76737,0.97183,0.95808,0.98712,0.87406,0.5981,0.34504,0.87076,0.001585,0.001585,0.001585
+87,16324.1,0.11337,0.10221,0.76695,0.97137,0.95881,0.98708,0.87382,0.59851,0.34519,0.87106,0.001486,0.001486,0.001486
+88,16517.1,0.11185,0.10043,0.76513,0.97121,0.95899,0.98707,0.87379,0.59906,0.34583,0.87135,0.001387,0.001387,0.001387
+89,16708.5,0.11103,0.09846,0.76565,0.97113,0.95904,0.98709,0.87369,0.59838,0.34599,0.87138,0.001288,0.001288,0.001288
+90,16896.9,0.11054,0.0982,0.76703,0.97095,0.95892,0.98712,0.87377,0.59757,0.34552,0.87126,0.001189,0.001189,0.001189
+91,17091.5,0.10967,0.09616,0.76665,0.97037,0.9595,0.98704,0.87361,0.59635,0.34561,0.87111,0.00109,0.00109,0.00109
+92,17282.9,0.10834,0.09481,0.76509,0.9726,0.95743,0.98704,0.87372,0.5956,0.34572,0.87108,0.000991,0.000991,0.000991
+93,17471.1,0.10692,0.09247,0.76461,0.97255,0.95738,0.9871,0.87368,0.59467,0.34689,0.8707,0.000892,0.000892,0.000892
+94,17654.7,0.10578,0.09076,0.76573,0.97167,0.95786,0.9872,0.87367,0.59367,0.34732,0.87049,0.000793,0.000793,0.000793
+95,17858.1,0.10457,0.08903,0.7648,0.97097,0.95816,0.98718,0.87394,0.59295,0.34757,0.87044,0.000694,0.000694,0.000694
+96,18048,0.10283,0.08802,0.76437,0.97358,0.95577,0.98712,0.8737,0.59392,0.34877,0.87087,0.000595,0.000595,0.000595
+97,18233,0.10269,0.08685,0.76468,0.97469,0.95492,0.98712,0.8741,0.59227,0.34903,0.87042,0.000496,0.000496,0.000496
+98,18418.2,0.10143,0.0852,0.7644,0.97473,0.95512,0.98709,0.87397,0.59171,0.35007,0.8704,0.000397,0.000397,0.000397
+99,18605.1,0.10052,0.08363,0.76442,0.97443,0.95526,0.98712,0.87396,0.5922,0.35121,0.87087,0.000298,0.000298,0.000298
+100,18790,0.09925,0.08228,0.76465,0.97498,0.95493,0.98711,0.8737,0.59312,0.35293,0.87138,0.000199,0.000199,0.000199
diff --git a/tests/shared/augmentation/__init__.py b/tests/shared/augmentation/__init__.py
new file mode 100644
index 0000000..56fc805
--- /dev/null
+++ b/tests/shared/augmentation/__init__.py
@@ -0,0 +1 @@
+# Tests for augmentation module
diff --git a/tests/shared/augmentation/test_base.py b/tests/shared/augmentation/test_base.py
new file mode 100644
index 0000000..3f238d1
--- /dev/null
+++ b/tests/shared/augmentation/test_base.py
@@ -0,0 +1,347 @@
+"""
+Tests for augmentation base module.
+
+TDD Phase 1: RED - Write tests first, then implement to pass.
+"""
+
+from typing import Any
+from unittest.mock import MagicMock
+
+import numpy as np
+import pytest
+
+
+class TestAugmentationResult:
+    """Tests for AugmentationResult dataclass."""
+
+    def test_minimal_result(self) -> None:
+        """Test creating result with only required field."""
+        from shared.augmentation.base import AugmentationResult
+
+        image = np.zeros((100, 100, 3), dtype=np.uint8)
+        result = AugmentationResult(image=image)
+
+        assert result.image is image
+        assert result.bboxes is None
+        assert result.transform_matrix is None
+        assert result.applied is True
+        assert result.metadata is None
+
+    def test_full_result(self) -> None:
+        """Test creating result with all fields."""
+        from shared.augmentation.base import AugmentationResult
+
+        image = np.zeros((100, 100, 3), dtype=np.uint8)
+        bboxes = np.array([[0, 0.5, 0.5, 0.1, 0.1]])
+        transform = np.eye(3)
+        metadata = {"applied_transform": "wrinkle"}
+
+        result = AugmentationResult(
+            image=image,
+            bboxes=bboxes,
+            transform_matrix=transform,
+            applied=True,
+            metadata=metadata,
+        )
+
+        assert result.image is image
+        np.testing.assert_array_equal(result.bboxes, bboxes)
+        np.testing.assert_array_equal(result.transform_matrix, transform)
+        assert result.applied is True
+        assert result.metadata == {"applied_transform": "wrinkle"}
+
+    def test_not_applied(self) -> None:
+        """Test result when augmentation was not applied."""
+        from shared.augmentation.base import AugmentationResult
+
+        image = np.zeros((100, 100, 3), dtype=np.uint8)
+        result = AugmentationResult(image=image, applied=False)
+
+        assert result.applied is False
+
+
+class TestBaseAugmentation:
+    """Tests for BaseAugmentation abstract class."""
+
+    def test_cannot_instantiate_directly(self) -> None:
+        """Test that BaseAugmentation cannot be instantiated."""
+        from shared.augmentation.base import BaseAugmentation
+
+        with pytest.raises(TypeError):
+            BaseAugmentation({})  # type: ignore
+
+    def test_subclass_must_implement_apply(self) -> None:
+        """Test that subclass must implement apply method."""
+        from shared.augmentation.base import BaseAugmentation
+
+        class IncompleteAugmentation(BaseAugmentation):
+            name = "incomplete"
+
+            def _validate_params(self) -> None:
+                pass
+
+            # Missing apply method
+
+        with pytest.raises(TypeError):
+            IncompleteAugmentation({})  # type: ignore
+
+    def test_subclass_must_implement_validate_params(self) -> None:
+        """Test that subclass must implement _validate_params."""
+        from shared.augmentation.base import AugmentationResult, BaseAugmentation
+
+        class IncompleteAugmentation(BaseAugmentation):
+            name = "incomplete"
+
+            def apply(
+                self,
+                image: np.ndarray,
+                bboxes: np.ndarray | None = None,
+                rng: np.random.Generator | None = None,
+            ) -> AugmentationResult:
+                return AugmentationResult(image=image)
+
+            # Missing _validate_params method
+
+        with pytest.raises(TypeError):
+            IncompleteAugmentation({})  # type: ignore
+
+    def test_valid_subclass(self) -> None:
+        """Test creating a valid subclass."""
+        from shared.augmentation.base import AugmentationResult, BaseAugmentation
+
+        class DummyAugmentation(BaseAugmentation):
+            name = "dummy"
+            affects_geometry = False
+
+            def _validate_params(self) -> None:
+                pass
+
+            def apply(
+                self,
+                image: np.ndarray,
+                bboxes: np.ndarray | None = None,
+                rng: np.random.Generator | None = None,
+            ) -> AugmentationResult:
+                return AugmentationResult(image=image, bboxes=bboxes)
+
+        aug = DummyAugmentation({"param1": "value1"})
+
+        assert aug.name == "dummy"
+        assert aug.affects_geometry is False
+        assert aug.params == {"param1": "value1"}
+
+    def test_apply_returns_augmentation_result(self) -> None:
+        """Test that apply returns AugmentationResult."""
+        from shared.augmentation.base import AugmentationResult, BaseAugmentation
+
+        class DummyAugmentation(BaseAugmentation):
+            name = "dummy"
+
+            def _validate_params(self) -> None:
+                pass
+
+            def apply(
+                self,
+                image: np.ndarray,
+                bboxes: np.ndarray | None = None,
+                rng: np.random.Generator | None = None,
+            ) -> AugmentationResult:
+                # Simple pass-through
+                return AugmentationResult(image=image, bboxes=bboxes)
+
+        aug = DummyAugmentation({})
+        image = np.zeros((100, 100, 3), dtype=np.uint8)
+        bboxes = np.array([[0, 0.5, 0.5, 0.1, 0.1]])
+
+        result = aug.apply(image, bboxes)
+
+        assert isinstance(result, AugmentationResult)
+        assert result.image is image
+        np.testing.assert_array_equal(result.bboxes, bboxes)
+
+    def test_affects_geometry_default(self) -> None:
+        """Test that affects_geometry defaults to False."""
+        from shared.augmentation.base import AugmentationResult, BaseAugmentation
+
+        class DummyAugmentation(BaseAugmentation):
+            name = "dummy"
+            # Not setting affects_geometry
+
+            def _validate_params(self) -> None:
+                pass
+
+            def apply(
+                self,
+                image: np.ndarray,
+                bboxes: np.ndarray | None = None,
+                rng: np.random.Generator | None = None,
+            ) -> AugmentationResult:
+                return AugmentationResult(image=image)
+
+        aug = DummyAugmentation({})
+
+        assert aug.affects_geometry is False
+
+    def test_validate_params_called_on_init(self) -> None:
+        """Test that _validate_params is called during initialization."""
+        from shared.augmentation.base import AugmentationResult, BaseAugmentation
+
+        validation_called = {"called": False}
+
+        class ValidatingAugmentation(BaseAugmentation):
+            name = "validating"
+
+            def _validate_params(self) -> None:
+                validation_called["called"] = True
+
+            def apply(
+                self,
+                image: np.ndarray,
+                bboxes: np.ndarray | None = None,
+                rng: np.random.Generator | None = None,
+            ) -> AugmentationResult:
+                return AugmentationResult(image=image)
+
+        ValidatingAugmentation({})
+
+        assert validation_called["called"] is True
+
+    def test_validate_params_raises_on_invalid(self) -> None:
+        """Test that _validate_params can raise ValueError."""
+        from shared.augmentation.base import AugmentationResult, BaseAugmentation
+
+        class StrictAugmentation(BaseAugmentation):
+            name = "strict"
+
+            def _validate_params(self) -> None:
+                if "required_param" not in self.params:
+                    raise ValueError("required_param is required")
+
+            def apply(
+                self,
+                image: np.ndarray,
+                bboxes: np.ndarray | None = None,
+                rng: np.random.Generator | None = None,
+            ) -> AugmentationResult:
+                return AugmentationResult(image=image)
+
+        with pytest.raises(ValueError, match="required_param"):
+            StrictAugmentation({})
+
+        # Should work with required param
+        aug = StrictAugmentation({"required_param": "value"})
+        assert aug.params["required_param"] == "value"
+
+    def test_rng_usage(self) -> None:
+        """Test that random generator can be passed and used."""
+        from shared.augmentation.base import AugmentationResult, BaseAugmentation
+
+        class RandomAugmentation(BaseAugmentation):
+            name = "random"
+
+            def _validate_params(self) -> None:
+                pass
+
+            def apply(
+                self,
+                image: np.ndarray,
+                bboxes: np.ndarray | None = None,
+                rng: np.random.Generator | None = None,
+            ) -> AugmentationResult:
+                if rng is None:
+                    rng = np.random.default_rng()
+                # Use rng to generate a random value
+                random_value = rng.random()
+                return AugmentationResult(
+                    image=image,
+                    metadata={"random_value": random_value},
+                )
+
+        aug = RandomAugmentation({})
+        image = np.zeros((100, 100, 3), dtype=np.uint8)
+
+        # With same seed, should get same result
+        rng1 = np.random.default_rng(42)
+        rng2 = np.random.default_rng(42)
+
+        result1 = aug.apply(image, rng=rng1)
+        result2 = aug.apply(image, rng=rng2)
+
+        assert result1.metadata is not None
+        assert result2.metadata is not None
+        assert result1.metadata["random_value"] == result2.metadata["random_value"]
+
+
+class TestAugmentationResultImmutability:
+    """Tests for ensuring result doesn't mutate input."""
+
+    def test_image_not_modified(self) -> None:
+        """Test that original image is not modified."""
+        from shared.augmentation.base import AugmentationResult, BaseAugmentation
+
+        class ModifyingAugmentation(BaseAugmentation):
+            name = "modifying"
+
+            def _validate_params(self) -> None:
+                pass
+
+            def apply(
+                self,
+                image: np.ndarray,
+                bboxes: np.ndarray | None = None,
+                rng: np.random.Generator | None = None,
+            ) -> AugmentationResult:
+                # Should copy before modifying
+                modified = image.copy()
+                modified[:] = 255
+                return AugmentationResult(image=modified)
+
+        aug = ModifyingAugmentation({})
+        original = np.zeros((100, 100, 3), dtype=np.uint8)
+        original_copy = original.copy()
+
+        result = aug.apply(original)
+
+        # Original should be unchanged
+        np.testing.assert_array_equal(original, original_copy)
+        # Result should be modified
+        assert np.all(result.image == 255)
+
+    def test_bboxes_not_modified(self) -> None:
+        """Test that original bboxes are not modified."""
+        from shared.augmentation.base import AugmentationResult, BaseAugmentation
+
+        class BboxModifyingAugmentation(BaseAugmentation):
+            name = "bbox_modifying"
+            affects_geometry = True
+
+            def _validate_params(self) -> None:
+                pass
+
+            def apply(
+                self,
+                image: np.ndarray,
+                bboxes: np.ndarray | None = None,
+                rng: np.random.Generator | None = None,
+            ) -> AugmentationResult:
+                if bboxes is not None:
+                    # Should copy before modifying
+                    modified_bboxes = bboxes.copy()
+                    modified_bboxes[:, 1:] *= 0.5  # Scale down
+                    return AugmentationResult(image=image, bboxes=modified_bboxes)
+                return AugmentationResult(image=image)
+
+        aug = BboxModifyingAugmentation({})
+        image = np.zeros((100, 100, 3), dtype=np.uint8)
+        original_bboxes = np.array([[0, 0.5, 0.5, 0.2, 0.2]], dtype=np.float32)
+        original_bboxes_copy = original_bboxes.copy()
+
+        result = aug.apply(image, original_bboxes)
+
+        # Original should be unchanged
+        np.testing.assert_array_equal(original_bboxes, original_bboxes_copy)
+        # Result should be modified
+        assert result.bboxes is not None
+        np.testing.assert_array_almost_equal(
+            result.bboxes, np.array([[0, 0.25, 0.25, 0.1, 0.1]])
+        )
diff --git a/tests/shared/augmentation/test_config.py b/tests/shared/augmentation/test_config.py
new file mode 100644
index 0000000..4b6047c
--- /dev/null
+++ b/tests/shared/augmentation/test_config.py
@@ -0,0 +1,283 @@
+"""
+Tests for augmentation configuration module.
+
+TDD Phase 1: RED - Write tests first, then implement to pass.
+"""
+
+from typing import Any
+
+import pytest
+
+
+class TestAugmentationParams:
+    """Tests for AugmentationParams dataclass."""
+
+    def test_default_values(self) -> None:
+        """Test default parameter values."""
+        from shared.augmentation.config import AugmentationParams
+
+        params = AugmentationParams()
+
+        assert params.enabled is False
+        assert params.probability == 0.5
+        assert params.params == {}
+
+    def test_custom_values(self) -> None:
+        """Test creating params with custom values."""
+        from shared.augmentation.config import AugmentationParams
+
+        params = AugmentationParams(
+            enabled=True,
+            probability=0.8,
+            params={"intensity": 0.5, "num_wrinkles": (2, 5)},
+        )
+
+        assert params.enabled is True
+        assert params.probability == 0.8
+        assert params.params["intensity"] == 0.5
+        assert params.params["num_wrinkles"] == (2, 5)
+
+    def test_immutability_params_dict(self) -> None:
+        """Test that params dict is independent between instances."""
+        from shared.augmentation.config import AugmentationParams
+
+        params1 = AugmentationParams()
+        params2 = AugmentationParams()
+
+        # Modifying one should not affect the other
+        params1.params["test"] = "value"
+
+        assert "test" not in params2.params
+
+    def test_to_dict(self) -> None:
+        """Test conversion to dictionary."""
+        from shared.augmentation.config import AugmentationParams
+
+        params = AugmentationParams(
+            enabled=True,
+            probability=0.7,
+            params={"key": "value"},
+        )
+
+        result = params.to_dict()
+
+        assert result == {
+            "enabled": True,
+            "probability": 0.7,
+            "params": {"key": "value"},
+        }
+
+    def test_from_dict(self) -> None:
+        """Test creation from dictionary."""
+        from shared.augmentation.config import AugmentationParams
+
+        data = {
+            "enabled": True,
+            "probability": 0.6,
+            "params": {"intensity": 0.3},
+        }
+
+        params = AugmentationParams.from_dict(data)
+
+        assert params.enabled is True
+        assert params.probability == 0.6
+        assert params.params == {"intensity": 0.3}
+
+    def test_from_dict_with_defaults(self) -> None:
+        """Test creation from partial dictionary uses defaults."""
+        from shared.augmentation.config import AugmentationParams
+
+        data: dict[str, Any] = {"enabled": True}
+
+        params = AugmentationParams.from_dict(data)
+
+        assert params.enabled is True
+        assert params.probability == 0.5  # default
+        assert params.params == {}  # default
+
+
+class TestAugmentationConfig:
+    """Tests for AugmentationConfig dataclass."""
+
+    def test_default_values(self) -> None:
+        """Test that all augmentation types have defaults."""
+        from shared.augmentation.config import AugmentationConfig
+
+        config = AugmentationConfig()
+
+        # All augmentation types should exist
+        augmentation_types = [
+            "perspective_warp",
+            "wrinkle",
+            "edge_damage",
+            "stain",
+            "lighting_variation",
+            "shadow",
+            "gaussian_blur",
+            "motion_blur",
+            "gaussian_noise",
+            "salt_pepper",
+            "paper_texture",
+            "scanner_artifacts",
+        ]
+
+        for aug_type in augmentation_types:
+            assert hasattr(config, aug_type), f"Missing augmentation type: {aug_type}"
+            params = getattr(config, aug_type)
+            assert hasattr(params, "enabled")
+            assert hasattr(params, "probability")
+            assert hasattr(params, "params")
+
+    def test_global_settings_defaults(self) -> None:
+        """Test global settings default values."""
+        from shared.augmentation.config import AugmentationConfig
+
+        config = AugmentationConfig()
+
+        assert config.preserve_bboxes is True
+        assert config.seed is None
+
+    def test_custom_seed(self) -> None:
+        """Test setting custom seed for reproducibility."""
+        from shared.augmentation.config import AugmentationConfig
+
+        config = AugmentationConfig(seed=42)
+
+        assert config.seed == 42
+
+    def test_to_dict(self) -> None:
+        """Test conversion to dictionary."""
+        from shared.augmentation.config import AugmentationConfig
+
+        config = AugmentationConfig(seed=123, preserve_bboxes=False)
+
+        result = config.to_dict()
+
+        assert isinstance(result, dict)
+        assert result["seed"] == 123
+        assert result["preserve_bboxes"] is False
+        assert "perspective_warp" in result
+        assert "wrinkle" in result
+
+    def test_from_dict(self) -> None:
+        """Test creation from dictionary."""
+        from shared.augmentation.config import AugmentationConfig
+
+        data = {
+            "seed": 456,
+            "preserve_bboxes": False,
+            "wrinkle": {
+                "enabled": True,
+                "probability": 0.8,
+                "params": {"intensity": 0.5},
+            },
+        }
+
+        config = AugmentationConfig.from_dict(data)
+
+        assert config.seed == 456
+        assert config.preserve_bboxes is False
+        assert config.wrinkle.enabled is True
+        assert config.wrinkle.probability == 0.8
+        assert config.wrinkle.params["intensity"] == 0.5
+
+    def test_from_dict_with_partial_data(self) -> None:
+        """Test creation from partial dictionary uses defaults."""
+        from shared.augmentation.config import AugmentationConfig
+
+        data: dict[str, Any] = {
+            "wrinkle": {"enabled": True},
+        }
+
+        config = AugmentationConfig.from_dict(data)
+
+        # Explicitly set value
+        assert config.wrinkle.enabled is True
+        # Default values
+        assert config.preserve_bboxes is True
+        assert config.seed is None
+        assert config.gaussian_blur.enabled is False
+
+    def test_get_enabled_augmentations(self) -> None:
+        """Test getting list of enabled augmentations."""
+        from shared.augmentation.config import AugmentationConfig, AugmentationParams
+
+        config = AugmentationConfig(
+            wrinkle=AugmentationParams(enabled=True),
+            stain=AugmentationParams(enabled=True),
+            gaussian_blur=AugmentationParams(enabled=False),
+        )
+
+        enabled = config.get_enabled_augmentations()
+
+        assert "wrinkle" in enabled
+        assert "stain" in enabled
+        assert "gaussian_blur" not in enabled
+
+    def test_document_safe_defaults(self) -> None:
+        """Test that default params are document-safe (conservative)."""
+        from shared.augmentation.config import AugmentationConfig
+
+        config = AugmentationConfig()
+
+        # Perspective warp should be very conservative
+        assert config.perspective_warp.params.get("max_warp", 0.02) <= 0.05
+
+        # Noise should be subtle
+        noise_std = config.gaussian_noise.params.get("std", (5, 15))
+        if isinstance(noise_std, tuple):
+            assert noise_std[1] <= 20  # Max std should be low
+
+    def test_immutability_between_instances(self) -> None:
+        """Test that config instances are independent."""
+        from shared.augmentation.config import AugmentationConfig
+
+        config1 = AugmentationConfig()
+        config2 = AugmentationConfig()
+
+        # Modifying one should not affect the other
+        config1.wrinkle.params["test"] = "value"
+
+        assert "test" not in config2.wrinkle.params
+
+
+class TestAugmentationConfigValidation:
+    """Tests for configuration validation."""
+
+    def test_probability_range_validation(self) -> None:
+        """Test that probability values are validated."""
+        from shared.augmentation.config import AugmentationParams
+
+        # Valid range
+        params = AugmentationParams(probability=0.5)
+        assert params.probability == 0.5
+
+        # Edge cases
+        params_zero = AugmentationParams(probability=0.0)
+        assert params_zero.probability == 0.0
+
+        params_one = AugmentationParams(probability=1.0)
+        assert params_one.probability == 1.0
+
+    def test_config_validate_method(self) -> None:
+        """Test the validate method catches invalid configurations."""
+        from shared.augmentation.config import AugmentationConfig, AugmentationParams
+
+        # Invalid probability
+        config = AugmentationConfig(
+            wrinkle=AugmentationParams(probability=1.5),  # Invalid
+        )
+
+        with pytest.raises(ValueError, match="probability"):
+            config.validate()
+
+    def test_config_validate_negative_probability(self) -> None:
+        """Test validation catches negative probability."""
+        from shared.augmentation.config import AugmentationConfig, AugmentationParams
+
+        config = AugmentationConfig(
+            wrinkle=AugmentationParams(probability=-0.1),
+        )
+
+        with pytest.raises(ValueError, match="probability"):
+            config.validate()
diff --git a/tests/shared/augmentation/test_pipeline.py b/tests/shared/augmentation/test_pipeline.py
new file mode 100644
index 0000000..f317b00
--- /dev/null
+++ b/tests/shared/augmentation/test_pipeline.py
@@ -0,0 +1,338 @@
+"""
+Tests for augmentation pipeline module.
+
+TDD Phase 2: RED - Write tests first, then implement to pass.
+"""
+
+from typing import Any
+from unittest.mock import MagicMock, patch
+
+import numpy as np
+import pytest
+
+
+class TestAugmentationPipeline:
+    """Tests for AugmentationPipeline class."""
+
+    def test_create_with_config(self) -> None:
+        """Test creating pipeline with config."""
+        from shared.augmentation.config import AugmentationConfig
+        from shared.augmentation.pipeline import AugmentationPipeline
+
+        config = AugmentationConfig()
+        pipeline = AugmentationPipeline(config)
+
+        assert pipeline.config is config
+
+    def test_create_with_seed(self) -> None:
+        """Test creating pipeline with seed for reproducibility."""
+        from shared.augmentation.config import AugmentationConfig
+        from shared.augmentation.pipeline import AugmentationPipeline
+
+        config = AugmentationConfig(seed=42)
+        pipeline = AugmentationPipeline(config)
+
+        assert pipeline.config.seed == 42
+
+    def test_apply_returns_augmentation_result(self) -> None:
+        """Test that apply returns AugmentationResult."""
+        from shared.augmentation.base import AugmentationResult
+        from shared.augmentation.config import AugmentationConfig
+        from shared.augmentation.pipeline import AugmentationPipeline
+
+        config = AugmentationConfig()
+        pipeline = AugmentationPipeline(config)
+
+        image = np.zeros((100, 100, 3), dtype=np.uint8)
+        result = pipeline.apply(image)
+
+        assert isinstance(result, AugmentationResult)
+        assert result.image is not None
+        assert result.image.shape == image.shape
+
+    def test_apply_with_bboxes(self) -> None:
+        """Test apply with bounding boxes."""
+        from shared.augmentation.config import AugmentationConfig
+        from shared.augmentation.pipeline import AugmentationPipeline
+
+        config = AugmentationConfig()
+        pipeline = AugmentationPipeline(config)
+
+        image = np.zeros((100, 100, 3), dtype=np.uint8)
+        bboxes = np.array([[0, 0.5, 0.5, 0.1, 0.1]], dtype=np.float32)
+
+        result = pipeline.apply(image, bboxes)
+
+        # Bboxes should be preserved when preserve_bboxes=True
+        assert result.bboxes is not None
+
+    def test_apply_no_augmentations_enabled(self) -> None:
+        """Test apply when no augmentations are enabled."""
+        from shared.augmentation.config import AugmentationConfig, AugmentationParams
+        from shared.augmentation.pipeline import AugmentationPipeline
+
+        # Disable all augmentations
+        config = AugmentationConfig(
+            lighting_variation=AugmentationParams(enabled=False),
+        )
+        pipeline = AugmentationPipeline(config)
+
+        image = np.random.randint(0, 255, (100, 100, 3), dtype=np.uint8)
+        result = pipeline.apply(image)
+
+        # Image should be unchanged (or a copy)
+        np.testing.assert_array_equal(result.image, image)
+
+    def test_apply_does_not_mutate_input(self) -> None:
+        """Test that apply does not mutate input image."""
+        from shared.augmentation.config import AugmentationConfig, AugmentationParams
+        from shared.augmentation.pipeline import AugmentationPipeline
+
+        config = AugmentationConfig(
+            lighting_variation=AugmentationParams(enabled=True, probability=1.0),
+        )
+        pipeline = AugmentationPipeline(config)
+
+        image = np.full((100, 100, 3), 128, dtype=np.uint8)
+        original_copy = image.copy()
+
+        pipeline.apply(image)
+
+        np.testing.assert_array_equal(image, original_copy)
+
+    def test_reproducibility_with_seed(self) -> None:
+        """Test that same seed produces same results."""
+        from shared.augmentation.config import AugmentationConfig, AugmentationParams
+        from shared.augmentation.pipeline import AugmentationPipeline
+
+        config1 = AugmentationConfig(
+            seed=42,
+            gaussian_noise=AugmentationParams(enabled=True, probability=1.0),
+        )
+        config2 = AugmentationConfig(
+            seed=42,
+            gaussian_noise=AugmentationParams(enabled=True, probability=1.0),
+        )
+
+        pipeline1 = AugmentationPipeline(config1)
+        pipeline2 = AugmentationPipeline(config2)
+
+        image = np.full((100, 100, 3), 128, dtype=np.uint8)
+
+        result1 = pipeline1.apply(image.copy())
+        result2 = pipeline2.apply(image.copy())
+
+        np.testing.assert_array_equal(result1.image, result2.image)
+
+    def test_metadata_contains_applied_augmentations(self) -> None:
+        """Test that metadata lists applied augmentations."""
+        from shared.augmentation.config import AugmentationConfig, AugmentationParams
+        from shared.augmentation.pipeline import AugmentationPipeline
+
+        config = AugmentationConfig(
+            seed=42,
+            gaussian_noise=AugmentationParams(enabled=True, probability=1.0),
+            lighting_variation=AugmentationParams(enabled=True, probability=1.0),
+        )
+        pipeline = AugmentationPipeline(config)
+
+        image = np.full((100, 100, 3), 128, dtype=np.uint8)
+        result = pipeline.apply(image)
+
+        assert result.metadata is not None
+        assert "applied_augmentations" in result.metadata
+        # Both should be applied with probability=1.0
+        assert "gaussian_noise" in result.metadata["applied_augmentations"]
+        assert "lighting_variation" in result.metadata["applied_augmentations"]
+
+
+class TestAugmentationPipelineStageOrder:
+    """Tests for pipeline stage ordering."""
+
+    def test_stage_order_defined(self) -> None:
+        """Test that stage order is defined."""
+        from shared.augmentation.pipeline import AugmentationPipeline
+
+        assert hasattr(AugmentationPipeline, "STAGE_ORDER")
+        expected_stages = [
+            "geometric",
+            "degradation",
+            "lighting",
+            "texture",
+            "blur",
+            "noise",
+        ]
+        assert AugmentationPipeline.STAGE_ORDER == expected_stages
+
+    def test_stage_mapping_defined(self) -> None:
+        """Test that all augmentation types are mapped to stages."""
+        from shared.augmentation.pipeline import AugmentationPipeline
+
+        assert hasattr(AugmentationPipeline, "STAGE_MAPPING")
+
+        expected_mappings = {
+            "perspective_warp": "geometric",
+            "wrinkle": "degradation",
+            "edge_damage": "degradation",
+            "stain": "degradation",
+            "lighting_variation": "lighting",
+            "shadow": "lighting",
+            "paper_texture": "texture",
+            "scanner_artifacts": "texture",
+            "gaussian_blur": "blur",
+            "motion_blur": "blur",
+            "gaussian_noise": "noise",
+            "salt_pepper": "noise",
+        }
+
+        for aug_name, stage in expected_mappings.items():
+            assert aug_name in AugmentationPipeline.STAGE_MAPPING
+            assert AugmentationPipeline.STAGE_MAPPING[aug_name] == stage
+
+    def test_geometric_before_degradation(self) -> None:
+        """Test that geometric transforms are applied before degradation."""
+        from shared.augmentation.pipeline import AugmentationPipeline
+
+        stages = AugmentationPipeline.STAGE_ORDER
+        geometric_idx = stages.index("geometric")
+        degradation_idx = stages.index("degradation")
+
+        assert geometric_idx < degradation_idx
+
+    def test_noise_applied_last(self) -> None:
+        """Test that noise is applied last."""
+        from shared.augmentation.pipeline import AugmentationPipeline
+
+        stages = AugmentationPipeline.STAGE_ORDER
+        assert stages[-1] == "noise"
+
+
+class TestAugmentationRegistry:
+    """Tests for augmentation registry."""
+
+    def test_registry_exists(self) -> None:
+        """Test that augmentation registry exists."""
+        from shared.augmentation.pipeline import AUGMENTATION_REGISTRY
+
+        assert isinstance(AUGMENTATION_REGISTRY, dict)
+
+    def test_registry_contains_all_types(self) -> None:
+        """Test that registry contains all augmentation types."""
+        from shared.augmentation.pipeline import AUGMENTATION_REGISTRY
+
+        expected_types = [
+            "perspective_warp",
+            "wrinkle",
+            "edge_damage",
+            "stain",
+            "lighting_variation",
+            "shadow",
+            "gaussian_blur",
+            "motion_blur",
+            "gaussian_noise",
+            "salt_pepper",
+            "paper_texture",
+            "scanner_artifacts",
+        ]
+
+        for aug_type in expected_types:
+            assert aug_type in AUGMENTATION_REGISTRY, f"Missing: {aug_type}"
+
+
+class TestPipelinePreview:
+    """Tests for pipeline preview functionality."""
+
+    def test_preview_single_augmentation(self) -> None:
+        """Test previewing a single augmentation."""
+        from shared.augmentation.config import AugmentationConfig, AugmentationParams
+        from shared.augmentation.pipeline import AugmentationPipeline
+
+        config = AugmentationConfig(
+            gaussian_noise=AugmentationParams(
+                enabled=True, probability=1.0, params={"std": (10, 10)}
+            ),
+        )
+        pipeline = AugmentationPipeline(config)
+
+        image = np.full((100, 100, 3), 128, dtype=np.uint8)
+        preview = pipeline.preview(image, "gaussian_noise")
+
+        assert preview.shape == image.shape
+        assert preview.dtype == np.uint8
+        # Preview should modify the image
+        assert not np.array_equal(preview, image)
+
+    def test_preview_unknown_augmentation_raises(self) -> None:
+        """Test that previewing unknown augmentation raises error."""
+        from shared.augmentation.config import AugmentationConfig
+        from shared.augmentation.pipeline import AugmentationPipeline
+
+        config = AugmentationConfig()
+        pipeline = AugmentationPipeline(config)
+
+        image = np.zeros((100, 100, 3), dtype=np.uint8)
+
+        with pytest.raises(ValueError, match="Unknown augmentation"):
+            pipeline.preview(image, "non_existent_augmentation")
+
+    def test_preview_is_deterministic(self) -> None:
+        """Test that preview produces deterministic results."""
+        from shared.augmentation.config import AugmentationConfig, AugmentationParams
+        from shared.augmentation.pipeline import AugmentationPipeline
+
+        config = AugmentationConfig(
+            gaussian_noise=AugmentationParams(enabled=True),
+        )
+        pipeline = AugmentationPipeline(config)
+
+        image = np.full((100, 100, 3), 128, dtype=np.uint8)
+
+        preview1 = pipeline.preview(image, "gaussian_noise")
+        preview2 = pipeline.preview(image, "gaussian_noise")
+
+        np.testing.assert_array_equal(preview1, preview2)
+
+
+class TestPipelineGetAvailableAugmentations:
+    """Tests for getting available augmentations."""
+
+    def test_get_available_augmentations(self) -> None:
+        """Test getting list of available augmentations."""
+        from shared.augmentation.pipeline import get_available_augmentations
+
+        augmentations = get_available_augmentations()
+
+        assert isinstance(augmentations, list)
+        assert len(augmentations) == 12
+
+        # Each item should have name, description, affects_geometry
+        for aug in augmentations:
+            assert "name" in aug
+            assert "description" in aug
+            assert "affects_geometry" in aug
+            assert "stage" in aug
+
+    def test_get_available_augmentations_includes_all_types(self) -> None:
+        """Test that all augmentation types are included."""
+        from shared.augmentation.pipeline import get_available_augmentations
+
+        augmentations = get_available_augmentations()
+        names = [aug["name"] for aug in augmentations]
+
+        expected = [
+            "perspective_warp",
+            "wrinkle",
+            "edge_damage",
+            "stain",
+            "lighting_variation",
+            "shadow",
+            "gaussian_blur",
+            "motion_blur",
+            "gaussian_noise",
+            "salt_pepper",
+            "paper_texture",
+            "scanner_artifacts",
+        ]
+
+        for name in expected:
+            assert name in names
diff --git a/tests/shared/augmentation/test_presets.py b/tests/shared/augmentation/test_presets.py
new file mode 100644
index 0000000..00e4051
--- /dev/null
+++ b/tests/shared/augmentation/test_presets.py
@@ -0,0 +1,102 @@
+"""
+Tests for augmentation presets module.
+
+TDD Phase 4: RED - Write tests first, then implement to pass.
+"""
+
+import pytest
+
+
+class TestPresets:
+    """Tests for augmentation presets."""
+
+    def test_presets_dict_exists(self) -> None:
+        """Test that PRESETS dictionary exists."""
+        from shared.augmentation.presets import PRESETS
+
+        assert isinstance(PRESETS, dict)
+        assert len(PRESETS) > 0
+
+    def test_expected_presets_exist(self) -> None:
+        """Test that expected presets are defined."""
+        from shared.augmentation.presets import PRESETS
+
+        expected_presets = ["conservative", "moderate", "aggressive", "scanned_document"]
+
+        for preset_name in expected_presets:
+            assert preset_name in PRESETS, f"Missing preset: {preset_name}"
+
+    def test_preset_structure(self) -> None:
+        """Test that each preset has required structure."""
+        from shared.augmentation.presets import PRESETS
+
+        for name, preset in PRESETS.items():
+            assert "description" in preset, f"Preset {name} missing description"
+            assert "config" in preset, f"Preset {name} missing config"
+            assert isinstance(preset["description"], str)
+            assert isinstance(preset["config"], dict)
+
+    def test_get_preset_config(self) -> None:
+        """Test getting config from preset."""
+        from shared.augmentation.presets import get_preset_config
+
+        config = get_preset_config("conservative")
+
+        assert config is not None
+        # Should have at least some augmentations defined
+        assert len(config) > 0
+
+    def test_get_preset_config_unknown_raises(self) -> None:
+        """Test that getting unknown preset raises error."""
+        from shared.augmentation.presets import get_preset_config
+
+        with pytest.raises(ValueError, match="Unknown preset"):
+            get_preset_config("nonexistent_preset")
+
+    def test_create_config_from_preset(self) -> None:
+        """Test creating AugmentationConfig from preset."""
+        from shared.augmentation.config import AugmentationConfig
+        from shared.augmentation.presets import create_config_from_preset
+
+        config = create_config_from_preset("moderate")
+
+        assert isinstance(config, AugmentationConfig)
+
+    def test_conservative_preset_is_safe(self) -> None:
+        """Test that conservative preset only enables safe augmentations."""
+        from shared.augmentation.presets import create_config_from_preset
+
+        config = create_config_from_preset("conservative")
+
+        # Should NOT enable geometric transforms
+        assert config.perspective_warp.enabled is False
+
+        # Should NOT enable heavy degradation
+        assert config.wrinkle.enabled is False
+        assert config.edge_damage.enabled is False
+        assert config.stain.enabled is False
+
+    def test_aggressive_preset_enables_more(self) -> None:
+        """Test that aggressive preset enables more augmentations."""
+        from shared.augmentation.presets import create_config_from_preset
+
+        config = create_config_from_preset("aggressive")
+
+        enabled = config.get_enabled_augmentations()
+
+        # Should enable multiple augmentation types
+        assert len(enabled) >= 6
+
+    def test_list_presets(self) -> None:
+        """Test listing available presets."""
+        from shared.augmentation.presets import list_presets
+
+        presets = list_presets()
+
+        assert isinstance(presets, list)
+        assert len(presets) >= 4
+
+        # Each item should have name and description
+        for preset in presets:
+            assert "name" in preset
+            assert "description" in preset
diff --git a/tests/shared/augmentation/transforms/__init__.py b/tests/shared/augmentation/transforms/__init__.py
new file mode 100644
index 0000000..f66e61f
--- /dev/null
+++ b/tests/shared/augmentation/transforms/__init__.py
@@ -0,0 +1 @@
+# Tests for augmentation transforms
diff --git a/tests/shared/test_dataset_augmenter.py b/tests/shared/test_dataset_augmenter.py
new file mode 100644
index 0000000..d9704e9
--- /dev/null
+++ b/tests/shared/test_dataset_augmenter.py
@@ -0,0 +1,293 @@
+"""
+Tests for DatasetAugmenter.
+
+TDD Phase 1: RED - Write tests first, then implement to pass.
+"""
+
+import tempfile
+from pathlib import Path
+
+import numpy as np
+import pytest
+from PIL import Image
+
+
+class TestDatasetAugmenter:
+    """Tests for DatasetAugmenter class."""
+
+    @pytest.fixture
+    def sample_dataset(self, tmp_path: Path) -> Path:
+        """Create a sample YOLO dataset structure."""
+        dataset_dir = tmp_path / "dataset"
+
+        # Create directory structure
+        for split in ["train", "val", "test"]:
+            (dataset_dir / "images" / split).mkdir(parents=True)
+            (dataset_dir / "labels" / split).mkdir(parents=True)
+
+        # Create sample images and labels
+        for i in range(3):
+            # Create 100x100 white image
+            img = Image.new("RGB", (100, 100), color="white")
+            img_path = dataset_dir / "images" / "train" / f"doc_{i}.png"
+            img.save(img_path)
+
+            # Create label with 2 bboxes
+            # Format: class_id x_center y_center width height
+            label_content = "0 0.5 0.3 0.2 0.1\n1 0.7 0.6 0.15 0.2\n"
+            label_path = dataset_dir / "labels" / "train" / f"doc_{i}.txt"
+            label_path.write_text(label_content)
+
+        # Create data.yaml
+        data_yaml = dataset_dir / "data.yaml"
+        data_yaml.write_text(
+            "path: .\n"
+            "train: images/train\n"
+            "val: images/val\n"
+            "test: images/test\n"
+            "nc: 10\n"
+            "names: [class0, class1, class2, class3, class4, class5, class6, class7, class8, class9]\n"
+        )
+
+        return dataset_dir
+
+    @pytest.fixture
+    def augmentation_config(self) -> dict:
+        """Create a sample augmentation config."""
+        return {
+            "gaussian_noise": {
+                "enabled": True,
+                "probability": 1.0,
+                "params": {"std": 10},
+            },
+            "gaussian_blur": {
+                "enabled": True,
+                "probability": 1.0,
+                "params": {"kernel_size": 3},
+            },
+        }
+
+    def test_augmenter_creates_additional_images(
+        self, sample_dataset: Path, augmentation_config: dict
+    ):
+        """Test that augmenter creates new augmented images."""
+        from shared.augmentation.dataset_augmenter import DatasetAugmenter
+
+        augmenter = DatasetAugmenter(augmentation_config)
+
+        # Count original images
+        original_count = len(list((sample_dataset / "images" / "train").glob("*.png")))
+        assert original_count == 3
+
+        # Apply augmentation with multiplier=2
+        result = augmenter.augment_dataset(sample_dataset, multiplier=2)
+
+        # Should now have original + 2x augmented = 3 + 6 = 9 images
+        new_count = len(list((sample_dataset / "images" / "train").glob("*.png")))
+        assert new_count == 9
+        assert result["augmented_images"] == 6
+
+    def test_augmenter_creates_matching_labels(
+        self, sample_dataset: Path, augmentation_config: dict
+    ):
+        """Test that augmenter creates label files for each augmented image."""
+        from shared.augmentation.dataset_augmenter import DatasetAugmenter
+
+        augmenter = DatasetAugmenter(augmentation_config)
+        augmenter.augment_dataset(sample_dataset, multiplier=2)
+
+        # Check that each image has a matching label file
+        images = list((sample_dataset / "images" / "train").glob("*.png"))
+        labels = list((sample_dataset / "labels" / "train").glob("*.txt"))
+
+        assert len(images) == len(labels)
+
+        # Check that augmented images have corresponding labels
+        for img_path in images:
+            label_path = sample_dataset / "labels" / "train" / f"{img_path.stem}.txt"
+            assert label_path.exists(), f"Missing label for {img_path.name}"
+
+    def test_augmented_labels_have_valid_format(
+        self, sample_dataset: Path, augmentation_config: dict
+    ):
+        """Test that augmented label files have valid YOLO format."""
+        from shared.augmentation.dataset_augmenter import DatasetAugmenter
+
+        augmenter = DatasetAugmenter(augmentation_config)
+        augmenter.augment_dataset(sample_dataset, multiplier=1)
+
+        # Check all label files
+        for label_path in (sample_dataset / "labels" / "train").glob("*.txt"):
+            content = label_path.read_text().strip()
+            if not content:
+                continue  # Empty labels are valid (background images)
+
+            for line in content.split("\n"):
+                parts = line.split()
+                assert len(parts) == 5, f"Invalid label format in {label_path.name}"
+
+                class_id = int(parts[0])
+                x_center = float(parts[1])
+                y_center = float(parts[2])
+                width = float(parts[3])
+                height = float(parts[4])
+
+                # Check values are in valid range
+                assert 0 <= class_id < 100, f"Invalid class_id: {class_id}"
+                assert 0 <= x_center <= 1, f"Invalid x_center: {x_center}"
+                assert 0 <= y_center <= 1, f"Invalid y_center: {y_center}"
+                assert 0 <= width <= 1, f"Invalid width: {width}"
+                assert 0 <= height <= 1, f"Invalid height: {height}"
+
+    def test_augmented_images_are_different(
+        self, sample_dataset: Path, augmentation_config: dict
+    ):
+        """Test that augmented images are actually different from originals."""
+        from shared.augmentation.dataset_augmenter import DatasetAugmenter
+
+        # Load original image
+        original_path = sample_dataset / "images" / "train" / "doc_0.png"
+        original_img = np.array(Image.open(original_path))
+
+        augmenter = DatasetAugmenter(augmentation_config)
+        augmenter.augment_dataset(sample_dataset, multiplier=1)
+
+        # Find augmented version
+        aug_path = sample_dataset / "images" / "train" / "doc_0_aug0.png"
+        assert aug_path.exists()
+
+        aug_img = np.array(Image.open(aug_path))
+
+        # Images should be different (due to noise/blur)
+        assert not np.array_equal(original_img, aug_img)
+
+    def test_augmented_images_same_size(
+        self, sample_dataset: Path, augmentation_config: dict
+    ):
+        """Test that augmented images have same size as originals."""
+        from shared.augmentation.dataset_augmenter import DatasetAugmenter
+
+        # Get original size
+        original_path = sample_dataset / "images" / "train" / "doc_0.png"
+        original_img = Image.open(original_path)
+        original_size = original_img.size
+
+        augmenter = DatasetAugmenter(augmentation_config)
+        augmenter.augment_dataset(sample_dataset, multiplier=1)
+
+        # Check all augmented images have same size
+        for img_path in (sample_dataset / "images" / "train").glob("*_aug*.png"):
+            img = Image.open(img_path)
+            assert img.size == original_size, f"{img_path.name} has wrong size"
+
+    def test_perspective_warp_updates_bboxes(self, sample_dataset: Path):
+        """Test that perspective_warp augmentation updates bbox coordinates."""
+        from shared.augmentation.dataset_augmenter import DatasetAugmenter
+
+        config = {
+            "perspective_warp": {
+                "enabled": True,
+                "probability": 1.0,
+                "params": {"max_warp": 0.05},  # Use larger warp for visible difference
+            },
+        }
+
+        # Read original label
+        original_label = (sample_dataset / "labels" / "train" / "doc_0.txt").read_text()
+        original_bboxes = [line.split() for line in original_label.strip().split("\n")]
+
+        augmenter = DatasetAugmenter(config)
+        augmenter.augment_dataset(sample_dataset, multiplier=1)
+
+        # Read augmented label
+        aug_label = (sample_dataset / "labels" / "train" / "doc_0_aug0.txt").read_text()
+        aug_bboxes = [line.split() for line in aug_label.strip().split("\n")]
+
+        # Same number of bboxes
+        assert len(original_bboxes) == len(aug_bboxes)
+
+        # At least one bbox should have different coordinates
+        # (perspective warp changes geometry)
+        differences_found = False
+        for orig, aug in zip(original_bboxes, aug_bboxes):
+            # Class ID should be same
+            assert orig[0] == aug[0]
+            # Coordinates might differ
+            if orig[1:] != aug[1:]:
+                differences_found = True
+
+        assert differences_found, "Perspective warp should change bbox coordinates"
+
+    def test_augmenter_only_processes_train_split(
+        self, sample_dataset: Path, augmentation_config: dict
+    ):
+        """Test that augmenter only processes train split by default."""
+        from shared.augmentation.dataset_augmenter import DatasetAugmenter
+
+        # Add a val image
+        val_img = Image.new("RGB", (100, 100), color="white")
+        val_img.save(sample_dataset / "images" / "val" / "val_doc.png")
+        (sample_dataset / "labels" / "val" / "val_doc.txt").write_text("0 0.5 0.5 0.1 0.1\n")
+
+        augmenter = DatasetAugmenter(augmentation_config)
+        augmenter.augment_dataset(sample_dataset, multiplier=2)
+
+        # Val should still have only 1 image
+        val_count = len(list((sample_dataset / "images" / "val").glob("*.png")))
+        assert val_count == 1
+
+    def test_augmenter_with_multiplier_zero_does_nothing(
+        self, sample_dataset: Path, augmentation_config: dict
+    ):
+        """Test that multiplier=0 creates no augmented images."""
+        from shared.augmentation.dataset_augmenter import DatasetAugmenter
+
+        original_count = len(list((sample_dataset / "images" / "train").glob("*.png")))
+
+        augmenter = DatasetAugmenter(augmentation_config)
+        result = augmenter.augment_dataset(sample_dataset, multiplier=0)
+
+        new_count = len(list((sample_dataset / "images" / "train").glob("*.png")))
+        assert new_count == original_count
+        assert result["augmented_images"] == 0
+
+    def test_augmenter_with_seed_is_reproducible(
+        self, sample_dataset: Path, augmentation_config: dict
+    ):
+        """Test that same seed produces same augmentation results."""
+        from shared.augmentation.dataset_augmenter import DatasetAugmenter
+
+        # Create two separate datasets
+        import shutil
+        dataset1 = sample_dataset
+        dataset2 = sample_dataset.parent / "dataset2"
+        shutil.copytree(dataset1, dataset2)
+
+        # Augment both with same seed
+        augmenter1 = DatasetAugmenter(augmentation_config, seed=42)
+        augmenter1.augment_dataset(dataset1, multiplier=1)
+
+        augmenter2 = DatasetAugmenter(augmentation_config, seed=42)
+        augmenter2.augment_dataset(dataset2, multiplier=1)
+
+        # Compare augmented images
+        aug1 = np.array(Image.open(dataset1 / "images" / "train" / "doc_0_aug0.png"))
+        aug2 = np.array(Image.open(dataset2 / "images" / "train" / "doc_0_aug0.png"))
+
+        assert np.array_equal(aug1, aug2), "Same seed should produce same augmentation"
+
+    def test_augmenter_returns_summary(
+        self, sample_dataset: Path, augmentation_config: dict
+    ):
+        """Test that augmenter returns a summary of what was done."""
+        from shared.augmentation.dataset_augmenter import DatasetAugmenter
+
+        augmenter = DatasetAugmenter(augmentation_config)
+        result = augmenter.augment_dataset(sample_dataset, multiplier=2)
+
+        assert "original_images" in result
+        assert "augmented_images" in result
+        assert "total_images" in result
+        assert result["original_images"] == 3
+        assert result["augmented_images"] == 6
+        assert result["total_images"] == 9
diff --git a/tests/web/test_augmentation_routes.py b/tests/web/test_augmentation_routes.py
new file mode 100644
index 0000000..698d876
--- /dev/null
+++ b/tests/web/test_augmentation_routes.py
@@ -0,0 +1,261 @@
+"""
+Tests for augmentation API routes.
+
+TDD Phase 5: RED - Write tests first, then implement to pass.
+"""
+
+import pytest
+from fastapi.testclient import TestClient
+
+
+class TestAugmentationTypesEndpoint:
+    """Tests for GET /admin/augmentation/types endpoint."""
+
+    def test_list_augmentation_types(
+        self, admin_client: TestClient, admin_token: str
+    ) -> None:
+        """Test listing available augmentation types."""
+        response = admin_client.get(
+            "/api/v1/admin/augmentation/types",
+            headers={"X-Admin-Token": admin_token},
+        )
+
+        assert response.status_code == 200
+        data = response.json()
+
+        assert "augmentation_types" in data
+        assert len(data["augmentation_types"]) == 12
+
+        # Check structure
+        aug_type = data["augmentation_types"][0]
+        assert "name" in aug_type
+        assert "description" in aug_type
+        assert "affects_geometry" in aug_type
+        assert "stage" in aug_type
+
+    def test_list_augmentation_types_unauthorized(
+        self, admin_client: TestClient
+    ) -> None:
+        """Test that unauthorized request is rejected."""
+        response = admin_client.get("/api/v1/admin/augmentation/types")
+
+        assert response.status_code == 401
+
+
+class TestAugmentationPresetsEndpoint:
+    """Tests for GET /admin/augmentation/presets endpoint."""
+
+    def test_list_presets(self, admin_client: TestClient, admin_token: str) -> None:
+        """Test listing available presets."""
+        response = admin_client.get(
+            "/api/v1/admin/augmentation/presets",
+            headers={"X-Admin-Token": admin_token},
+        )
+
+        assert response.status_code == 200
+        data = response.json()
+
+        assert "presets" in data
+        assert len(data["presets"]) >= 4
+
+        # Check expected presets exist
+        preset_names = [p["name"] for p in data["presets"]]
+        assert "conservative" in preset_names
+        assert "moderate" in preset_names
+        assert "aggressive" in preset_names
+        assert "scanned_document" in preset_names
+
+
+class TestAugmentationPreviewEndpoint:
+    """Tests for POST /admin/augmentation/preview/{document_id} endpoint."""
+
+    def test_preview_augmentation(
+        self,
+        admin_client: TestClient,
+        admin_token: str,
+        sample_document_id: str,
+    ) -> None:
+        """Test previewing augmentation on a document."""
+        response = admin_client.post(
+            f"/api/v1/admin/augmentation/preview/{sample_document_id}",
+            headers={"X-Admin-Token": admin_token},
+            json={
+                "augmentation_type": "gaussian_noise",
+                "params": {"std": 15},
+            },
+        )
+
+        assert response.status_code == 200
+        data = response.json()
+
+        assert "preview_url" in data
+        assert "original_url" in data
+        assert "applied_params" in data
+
+    def test_preview_invalid_augmentation_type(
+        self,
+        admin_client: TestClient,
+        admin_token: str,
+        sample_document_id: str,
+    ) -> None:
+        """Test that invalid augmentation type returns error."""
+        response = admin_client.post(
+            f"/api/v1/admin/augmentation/preview/{sample_document_id}",
+            headers={"X-Admin-Token": admin_token},
+            json={
+                "augmentation_type": "nonexistent",
+                "params": {},
+            },
+        )
+
+        assert response.status_code == 400
+
+    def test_preview_nonexistent_document(
+        self,
+        admin_client: TestClient,
+        admin_token: str,
+    ) -> None:
+        """Test that nonexistent document returns 404."""
+        response = admin_client.post(
+            "/api/v1/admin/augmentation/preview/00000000-0000-0000-0000-000000000000",
+            headers={"X-Admin-Token": admin_token},
+            json={
+                "augmentation_type": "gaussian_noise",
+                "params": {},
+            },
+        )
+
+        assert response.status_code == 404
+
+
+class TestAugmentationPreviewConfigEndpoint:
+    """Tests for POST /admin/augmentation/preview-config/{document_id} endpoint."""
+
+    def test_preview_config(
+        self,
+        admin_client: TestClient,
+        admin_token: str,
+        sample_document_id: str,
+    ) -> None:
+        """Test previewing full config on a document."""
+        response = admin_client.post(
+            f"/api/v1/admin/augmentation/preview-config/{sample_document_id}",
+            headers={"X-Admin-Token": admin_token},
+            json={
+                "gaussian_noise": {"enabled": True, "probability": 1.0},
+                "lighting_variation": {"enabled": True, "probability": 1.0},
+                "preserve_bboxes": True,
+                "seed": 42,
+            },
+        )
+
+        assert response.status_code == 200
+        data = response.json()
+
+        assert "preview_url" in data
+        assert "original_url" in data
+
+
+class TestAugmentationBatchEndpoint:
+    """Tests for POST /admin/augmentation/batch endpoint."""
+
+    def test_create_augmented_dataset(
+        self,
+        admin_client: TestClient,
+        admin_token: str,
+        sample_dataset_id: str,
+    ) -> None:
+        """Test creating augmented dataset."""
+        response = admin_client.post(
+            "/api/v1/admin/augmentation/batch",
+            headers={"X-Admin-Token": admin_token},
+            json={
+                "dataset_id": sample_dataset_id,
+                "config": {
+                    "gaussian_noise": {"enabled": True, "probability": 0.5},
+                    "preserve_bboxes": True,
+                },
+                "output_name": "test_augmented_dataset",
+                "multiplier": 2,
+            },
+        )
+
+        assert response.status_code == 200
+        data = response.json()
+
+        assert "task_id" in data
+        assert "status" in data
+        assert "estimated_images" in data
+
+    def test_create_augmented_dataset_invalid_multiplier(
+        self,
+        admin_client: TestClient,
+        admin_token: str,
+        sample_dataset_id: str,
+    ) -> None:
+        """Test that invalid multiplier is rejected."""
+        response = admin_client.post(
+            "/api/v1/admin/augmentation/batch",
+            headers={"X-Admin-Token": admin_token},
+            json={
+                "dataset_id": sample_dataset_id,
+                "config": {},
+                "output_name": "test",
+                "multiplier": 100,  # Too high
+            },
+        )
+
+        assert response.status_code == 422  # Validation error
+
+
+class TestAugmentedDatasetsListEndpoint:
+    """Tests for GET /admin/augmentation/datasets endpoint."""
+
+    def test_list_augmented_datasets(
+        self, admin_client: TestClient, admin_token: str
+    ) -> None:
+        """Test listing augmented datasets."""
+        response = admin_client.get(
+            "/api/v1/admin/augmentation/datasets",
+            headers={"X-Admin-Token": admin_token},
+        )
+
+        assert response.status_code == 200
+        data = response.json()
+
+        assert "total" in data
+        assert "limit" in data
+        assert "offset" in data
+        assert "datasets" in data
+        assert isinstance(data["datasets"], list)
+
+    def test_list_augmented_datasets_pagination(
+        self, admin_client: TestClient, admin_token: str
+    ) -> None:
+        """Test pagination parameters."""
+        response = admin_client.get(
+            "/api/v1/admin/augmentation/datasets",
+            headers={"X-Admin-Token": admin_token},
+            params={"limit": 5, "offset": 0},
+        )
+
+        assert response.status_code == 200
+        data = response.json()
+
+        assert data["limit"] == 5
+        assert data["offset"] == 0
+
+
+# Fixtures for tests
+@pytest.fixture
+def sample_document_id() -> str:
+    """Provide a sample document ID for testing."""
+    # This would need to be created in test setup
+    return "test-document-id"
+
+
+@pytest.fixture
+def sample_dataset_id() -> str:
+    """Provide a sample dataset ID for testing."""
+    # This would need to be created in test setup
+    return "test-dataset-id"
diff --git a/tests/web/test_dataset_builder.py b/tests/web/test_dataset_builder.py
index ae79912..1c052d4 100644
--- a/tests/web/test_dataset_builder.py
+++ b/tests/web/test_dataset_builder.py
@@ -329,3 +329,414 @@ class TestDatasetBuilder:
             results.append([(d["document_id"], d["split"]) for d in docs])
 
         assert results[0] == results[1]
+
+
+class TestAssignSplitsByGroup:
+    """Tests for _assign_splits_by_group method with group_key logic."""
+
+    def _make_mock_doc(self, doc_id, group_key=None):
+        """Create a mock AdminDocument with document_id and group_key."""
+        doc = MagicMock(spec=AdminDocument)
+        doc.document_id = doc_id
+        doc.group_key = group_key
+        doc.page_count = 1
+        return doc
+
+    def test_single_doc_groups_are_distributed(self, tmp_path, mock_admin_db):
+        """Documents with unique group_key are distributed across splits."""
+        from inference.web.services.dataset_builder import DatasetBuilder
+
+        builder = DatasetBuilder(db=mock_admin_db, base_dir=tmp_path / "datasets")
+
+        # 3 documents, each with unique group_key
+        docs = [
+            self._make_mock_doc(uuid4(), group_key="group-A"),
+            self._make_mock_doc(uuid4(), group_key="group-B"),
+            self._make_mock_doc(uuid4(), group_key="group-C"),
+        ]
+
+        result = builder._assign_splits_by_group(docs, train_ratio=0.7, val_ratio=0.2, seed=42)
+
+        # With 3 groups: 70% train = 2, 20% val = 1 (at least 1)
+        train_count = sum(1 for s in result.values() if s == "train")
+        val_count = sum(1 for s in result.values() if s == "val")
+        assert train_count >= 1
+        assert val_count >= 1  # Ensure val is not empty
+
+    def test_null_group_key_treated_as_single_doc_group(self, tmp_path, mock_admin_db):
+        """Documents with null/empty group_key are each treated as independent single-doc groups."""
+        from inference.web.services.dataset_builder import DatasetBuilder
+
+        builder = DatasetBuilder(db=mock_admin_db, base_dir=tmp_path / "datasets")
+
+        docs = [
+            self._make_mock_doc(uuid4(), group_key=None),
+            self._make_mock_doc(uuid4(), group_key=""),
+            self._make_mock_doc(uuid4(), group_key=None),
+        ]
+
+        result = builder._assign_splits_by_group(docs, train_ratio=0.7, val_ratio=0.2, seed=42)
+
+        # Each null/empty group_key doc is independent, distributed across splits
+        # With 3 docs: ensure at least 1 in train and 1 in val
+        train_count = sum(1 for s in result.values() if s == "train")
+        val_count = sum(1 for s in result.values() if s == "val")
+        assert train_count >= 1
+        assert val_count >= 1
+
+    def test_multi_doc_groups_stay_together(self, tmp_path, mock_admin_db):
+        """Documents with same group_key should be assigned to the same split."""
+        from inference.web.services.dataset_builder import DatasetBuilder
+
+        builder = DatasetBuilder(db=mock_admin_db, base_dir=tmp_path / "datasets")
+
+        # 6 documents in 2 groups
+        docs = [
+            self._make_mock_doc(uuid4(), group_key="supplier-A"),
+            self._make_mock_doc(uuid4(), group_key="supplier-A"),
+            self._make_mock_doc(uuid4(), group_key="supplier-A"),
+            self._make_mock_doc(uuid4(), group_key="supplier-B"),
+            self._make_mock_doc(uuid4(), group_key="supplier-B"),
+            self._make_mock_doc(uuid4(), group_key="supplier-B"),
+        ]
+
+        result = builder._assign_splits_by_group(docs, train_ratio=0.5, val_ratio=0.5, seed=42)
+
+        # All docs in supplier-A should have same split
+        splits_a = [result[str(d.document_id)] for d in docs[:3]]
+        assert len(set(splits_a)) == 1, "All docs in supplier-A should be in same split"
+
+        # All docs in supplier-B should have same split
+        splits_b = [result[str(d.document_id)] for d in docs[3:]]
+        assert len(set(splits_b)) == 1, "All docs in supplier-B should be in same split"
+
+    def test_multi_doc_groups_split_by_ratio(self, tmp_path, mock_admin_db):
+        """Multi-doc groups should be split according to train/val/test ratios."""
+        from inference.web.services.dataset_builder import DatasetBuilder
+
+        builder = DatasetBuilder(db=mock_admin_db, base_dir=tmp_path / "datasets")
+
+        # 10 groups with 2 docs each
+        docs = []
+        for i in range(10):
+            group_key = f"group-{i}"
+            docs.append(self._make_mock_doc(uuid4(), group_key=group_key))
+            docs.append(self._make_mock_doc(uuid4(), group_key=group_key))
+
+        result = builder._assign_splits_by_group(docs, train_ratio=0.7, val_ratio=0.2, seed=42)
+
+        # Count groups per split
+        group_splits = {}
+        for doc in docs:
+            split = result[str(doc.document_id)]
+            if doc.group_key not in group_splits:
+                group_splits[doc.group_key] = split
+            else:
+                # Verify same group has same split
+                assert group_splits[doc.group_key] == split
+
+        split_counts = {"train": 0, "val": 0, "test": 0}
+        for split in group_splits.values():
+            split_counts[split] += 1
+
+        # With 10 groups, 70/20/10 -> ~7 train, ~2 val, ~1 test
+        assert split_counts["train"] >= 6
+        assert split_counts["train"] <= 8
+        assert split_counts["val"] >= 1
+        assert split_counts["val"] <= 3
+
+    def test_mixed_single_and_multi_doc_groups(self, tmp_path, mock_admin_db):
+        """Mix of single-doc and multi-doc groups should be handled correctly."""
+        from inference.web.services.dataset_builder import DatasetBuilder
+
+        builder = DatasetBuilder(db=mock_admin_db, base_dir=tmp_path / "datasets")
+
+        docs = [
+            # Single-doc groups
+            self._make_mock_doc(uuid4(), group_key="single-1"),
+            self._make_mock_doc(uuid4(), group_key="single-2"),
+            self._make_mock_doc(uuid4(), group_key=None),
+            # Multi-doc groups
+            self._make_mock_doc(uuid4(), group_key="multi-A"),
+            self._make_mock_doc(uuid4(), group_key="multi-A"),
+            self._make_mock_doc(uuid4(), group_key="multi-B"),
+            self._make_mock_doc(uuid4(), group_key="multi-B"),
+        ]
+
+        result = builder._assign_splits_by_group(docs, train_ratio=0.5, val_ratio=0.5, seed=42)
+
+        # All groups are shuffled and distributed
+        # Ensure at least 1 in train and 1 in val
+        train_count = sum(1 for s in result.values() if s == "train")
+        val_count = sum(1 for s in result.values() if s == "val")
+        assert train_count >= 1
+        assert val_count >= 1
+
+        # Multi-doc groups stay together
+        assert result[str(docs[3].document_id)] == result[str(docs[4].document_id)]
+        assert result[str(docs[5].document_id)] == result[str(docs[6].document_id)]
+
+    def test_deterministic_with_seed(self, tmp_path, mock_admin_db):
+        """Same seed should produce same split assignments."""
+        from inference.web.services.dataset_builder import DatasetBuilder
+
+        builder = DatasetBuilder(db=mock_admin_db, base_dir=tmp_path / "datasets")
+
+        docs = [
+            self._make_mock_doc(uuid4(), group_key="group-A"),
+            self._make_mock_doc(uuid4(), group_key="group-A"),
+            self._make_mock_doc(uuid4(), group_key="group-B"),
+            self._make_mock_doc(uuid4(), group_key="group-B"),
+            self._make_mock_doc(uuid4(), group_key="group-C"),
+            self._make_mock_doc(uuid4(), group_key="group-C"),
+        ]
+
+        result1 = builder._assign_splits_by_group(docs, train_ratio=0.5, val_ratio=0.3, seed=123)
+        result2 = builder._assign_splits_by_group(docs, train_ratio=0.5, val_ratio=0.3, seed=123)
+
+        assert result1 == result2
+
+    def test_different_seed_may_produce_different_splits(self, tmp_path, mock_admin_db):
+        """Different seeds should potentially produce different split assignments."""
+        from inference.web.services.dataset_builder import DatasetBuilder
+
+        builder = DatasetBuilder(db=mock_admin_db, base_dir=tmp_path / "datasets")
+
+        # Many groups to increase chance of different results
+        docs = []
+        for i in range(20):
+            group_key = f"group-{i}"
+            docs.append(self._make_mock_doc(uuid4(), group_key=group_key))
+            docs.append(self._make_mock_doc(uuid4(), group_key=group_key))
+
+        result1 = builder._assign_splits_by_group(docs, train_ratio=0.5, val_ratio=0.3, seed=1)
+        result2 = builder._assign_splits_by_group(docs, train_ratio=0.5, val_ratio=0.3, seed=999)
+
+        # Results should be different (very likely with 20 groups)
+        assert result1 != result2
+
+    def test_all_docs_assigned(self, tmp_path, mock_admin_db):
+        """Every document should be assigned a split."""
+        from inference.web.services.dataset_builder import DatasetBuilder
+
+        builder = DatasetBuilder(db=mock_admin_db, base_dir=tmp_path / "datasets")
+
+        docs = [
+            self._make_mock_doc(uuid4(), group_key="group-A"),
+            self._make_mock_doc(uuid4(), group_key="group-A"),
+            self._make_mock_doc(uuid4(), group_key=None),
+            self._make_mock_doc(uuid4(), group_key="single"),
+        ]
+
+        result = builder._assign_splits_by_group(docs, train_ratio=0.7, val_ratio=0.2, seed=42)
+
+        assert len(result) == len(docs)
+        for doc in docs:
+            assert str(doc.document_id) in result
+            assert result[str(doc.document_id)] in ["train", "val", "test"]
+
+    def test_empty_documents_list(self, tmp_path, mock_admin_db):
+        """Empty document list should return empty result."""
+        from inference.web.services.dataset_builder import DatasetBuilder
+
+        builder = DatasetBuilder(db=mock_admin_db, base_dir=tmp_path / "datasets")
+
+        result = builder._assign_splits_by_group([], train_ratio=0.7, val_ratio=0.2, seed=42)
+
+        assert result == {}
+
+    def test_only_multi_doc_groups(self, tmp_path, mock_admin_db):
+        """When all groups have multiple docs, splits should follow ratios."""
+        from inference.web.services.dataset_builder import DatasetBuilder
+
+        builder = DatasetBuilder(db=mock_admin_db, base_dir=tmp_path / "datasets")
+
+        # 5 groups with 3 docs each
+        docs = []
+        for i in range(5):
+            group_key = f"group-{i}"
+            for _ in range(3):
+                docs.append(self._make_mock_doc(uuid4(), group_key=group_key))
+
+        result = builder._assign_splits_by_group(docs, train_ratio=0.6, val_ratio=0.2, seed=42)
+
+        # Group splits
+        group_splits = {}
+        for doc in docs:
+            if doc.group_key not in group_splits:
+                group_splits[doc.group_key] = result[str(doc.document_id)]
+
+        split_counts = {"train": 0, "val": 0, "test": 0}
+        for split in group_splits.values():
+            split_counts[split] += 1
+
+        # With 5 groups, 60/20/20 -> 3 train, 1 val, 1 test
+        assert split_counts["train"] >= 2
+        assert split_counts["train"] <= 4
+
+    def test_only_single_doc_groups(self, tmp_path, mock_admin_db):
+        """When all groups have single doc, they are distributed across splits."""
+        from inference.web.services.dataset_builder import DatasetBuilder
+
+        builder = DatasetBuilder(db=mock_admin_db, base_dir=tmp_path / "datasets")
+
+        docs = [
+            self._make_mock_doc(uuid4(), group_key="unique-1"),
+            self._make_mock_doc(uuid4(), group_key="unique-2"),
+            self._make_mock_doc(uuid4(), group_key="unique-3"),
+            self._make_mock_doc(uuid4(), group_key=None),
+            self._make_mock_doc(uuid4(), group_key=""),
+        ]
+
+        result = builder._assign_splits_by_group(docs, train_ratio=0.6, val_ratio=0.2, seed=42)
+
+        # With 5 groups: 60% train = 3, 20% val = 1 (at least 1)
+        train_count = sum(1 for s in result.values() if s == "train")
+        val_count = sum(1 for s in result.values() if s == "val")
+        assert train_count >= 2
+        assert val_count >= 1  # Ensure val is not empty
+
+
+class TestBuildDatasetWithGroupKey:
+    """Integration tests for build_dataset with group_key logic."""
+
+    @pytest.fixture
+    def grouped_documents(self, tmp_path):
+        """Create documents with various group_key configurations."""
+        doc_ids = []
+        docs = []
+
+        # Create 3 groups: 2 multi-doc groups + 2 single-doc groups
+        group_configs = [
+            ("supplier-A", 3),  # Multi-doc group: 3 docs
+            ("supplier-B", 2),  # Multi-doc group: 2 docs
+            ("unique-1", 1),    # Single-doc group
+            (None, 1),          # Null group_key
+        ]
+
+        for group_key, count in group_configs:
+            for _ in range(count):
+                doc_id = uuid4()
+                doc_ids.append(doc_id)
+
+                # Create image files
+                doc_dir = tmp_path / "admin_images" / str(doc_id)
+                doc_dir.mkdir(parents=True)
+                for page in range(1, 3):
+                    (doc_dir / f"page_{page}.png").write_bytes(b"fake-png")
+
+                # Create mock document
+                doc = MagicMock(spec=AdminDocument)
+                doc.document_id = doc_id
+                doc.filename = f"{doc_id}.pdf"
+                doc.page_count = 2
+                doc.group_key = group_key
+                doc.file_path = str(doc_dir)
+                docs.append(doc)
+
+        return tmp_path, docs
+
+    @pytest.fixture
+    def grouped_annotations(self, grouped_documents):
+        """Create annotations for grouped documents."""
+        tmp_path, docs = grouped_documents
+        annotations = {}
+        for doc in docs:
+            doc_anns = []
+            for page in range(1, 3):
+                ann = MagicMock(spec=AdminAnnotation)
+                ann.document_id = doc.document_id
+                ann.page_number = page
+                ann.class_id = 0
+                ann.class_name = "invoice_number"
+                ann.x_center = 0.5
+                ann.y_center = 0.3
+                ann.width = 0.2
+                ann.height = 0.05
+                doc_anns.append(ann)
+            annotations[str(doc.document_id)] = doc_anns
+        return annotations
+
+    def test_build_respects_group_key_splits(
+        self, grouped_documents, grouped_annotations, mock_admin_db
+    ):
+        """build_dataset should use group_key for split assignment."""
+        from inference.web.services.dataset_builder import DatasetBuilder
+
+        tmp_path, docs = grouped_documents
+
+        builder = DatasetBuilder(db=mock_admin_db, base_dir=tmp_path / "datasets")
+        mock_admin_db.get_documents_by_ids.return_value = docs
+        mock_admin_db.get_annotations_for_document.side_effect = lambda doc_id: (
+            grouped_annotations.get(str(doc_id), [])
+        )
+
+        dataset = mock_admin_db.create_dataset.return_value
+        builder.build_dataset(
+            dataset_id=str(dataset.dataset_id),
+            document_ids=[str(d.document_id) for d in docs],
+            train_ratio=0.5,
+            val_ratio=0.5,
+            seed=42,
+            admin_images_dir=tmp_path / "admin_images",
+        )
+
+        # Get the document splits from add_dataset_documents call
+        call_args = mock_admin_db.add_dataset_documents.call_args
+        docs_added = call_args[1]["documents"] if "documents" in call_args[1] else call_args[0][1]
+
+        # Build mapping of doc_id -> split
+        doc_split_map = {d["document_id"]: d["split"] for d in docs_added}
+
+        # Verify all docs are assigned a valid split
+        for doc_id in doc_split_map:
+            assert doc_split_map[doc_id] in ("train", "val", "test")
+
+        # Verify multi-doc groups stay together
+        supplier_a_ids = [str(d.document_id) for d in docs if d.group_key == "supplier-A"]
+        supplier_a_splits = [doc_split_map[doc_id] for doc_id in supplier_a_ids]
+        assert len(set(supplier_a_splits)) == 1, "supplier-A docs should be in same split"
+
+        supplier_b_ids = [str(d.document_id) for d in docs if d.group_key == "supplier-B"]
+        supplier_b_splits = [doc_split_map[doc_id] for doc_id in supplier_b_ids]
+        assert len(set(supplier_b_splits)) == 1, "supplier-B docs should be in same split"
+
+    def test_build_with_all_same_group_key(self, tmp_path, mock_admin_db):
+        """All docs with same group_key should go to same split."""
+        from inference.web.services.dataset_builder import DatasetBuilder
+
+        # Create 5 docs all with same group_key
+        docs = []
+        for i in range(5):
+            doc_id = uuid4()
+            doc_dir = tmp_path / "admin_images" / str(doc_id)
+            doc_dir.mkdir(parents=True)
+            (doc_dir / "page_1.png").write_bytes(b"fake-png")
+
+            doc = MagicMock(spec=AdminDocument)
+            doc.document_id = doc_id
+            doc.filename = f"{doc_id}.pdf"
+            doc.page_count = 1
+            doc.group_key = "same-group"
+            docs.append(doc)
+
+        builder = DatasetBuilder(db=mock_admin_db, base_dir=tmp_path / "datasets")
+        mock_admin_db.get_documents_by_ids.return_value = docs
+        mock_admin_db.get_annotations_for_document.return_value = []
+
+        dataset = mock_admin_db.create_dataset.return_value
+        builder.build_dataset(
+            dataset_id=str(dataset.dataset_id),
+            document_ids=[str(d.document_id) for d in docs],
+            train_ratio=0.6,
+            val_ratio=0.2,
+            seed=42,
+            admin_images_dir=tmp_path / "admin_images",
+        )
+
+        call_args = mock_admin_db.add_dataset_documents.call_args
+        docs_added = call_args[1]["documents"] if "documents" in call_args[1] else call_args[0][1]
+
+        splits = [d["split"] for d in docs_added]
+        # All should be in the same split (one group)
+        assert len(set(splits)) == 1, "All docs with same group_key should be in same split"
diff --git a/tests/web/test_dataset_routes.py b/tests/web/test_dataset_routes.py
index d2add37..2f1e5a5 100644
--- a/tests/web/test_dataset_routes.py
+++ b/tests/web/test_dataset_routes.py
@@ -25,6 +25,9 @@ TEST_DOC_UUID_2 = "990e8400-e29b-41d4-a716-446655440012"
 TEST_TOKEN = "test-admin-token-12345"
 TEST_TASK_UUID = "770e8400-e29b-41d4-a716-446655440002"
 
+# Generate 10 unique UUIDs for minimum document count tests
+TEST_DOC_UUIDS = [f"990e8400-e29b-41d4-a716-4466554400{i:02d}" for i in range(10, 20)]
+
 
 def _make_dataset(**overrides) -> MagicMock:
     defaults = dict(
@@ -83,14 +86,14 @@ class TestCreateDatasetRoute:
 
         mock_builder = MagicMock()
         mock_builder.build_dataset.return_value = {
-            "total_documents": 2,
-            "total_images": 4,
-            "total_annotations": 10,
+            "total_documents": 10,
+            "total_images": 20,
+            "total_annotations": 50,
         }
 
         request = DatasetCreateRequest(
             name="test-dataset",
-            document_ids=[TEST_DOC_UUID_1, TEST_DOC_UUID_2],
+            document_ids=TEST_DOC_UUIDS,  # Use 10 documents to meet minimum
         )
 
         with patch(
@@ -104,6 +107,73 @@ class TestCreateDatasetRoute:
         assert result.dataset_id == TEST_DATASET_UUID
         assert result.name == "test-dataset"
 
+    def test_create_dataset_fails_with_less_than_10_documents(self):
+        """Test that creating dataset fails if fewer than 10 documents provided."""
+        fn = _find_endpoint("create_dataset")
+
+        mock_db = MagicMock()
+
+        # Only 2 documents - should fail
+        request = DatasetCreateRequest(
+            name="test-dataset",
+            document_ids=[TEST_DOC_UUID_1, TEST_DOC_UUID_2],
+        )
+
+        from fastapi import HTTPException
+
+        with pytest.raises(HTTPException) as exc_info:
+            asyncio.run(fn(request=request, admin_token=TEST_TOKEN, db=mock_db))
+
+        assert exc_info.value.status_code == 400
+        assert "Minimum 10 documents required" in exc_info.value.detail
+        assert "got 2" in exc_info.value.detail
+        # Ensure DB was never called since validation failed first
+        mock_db.create_dataset.assert_not_called()
+
+    def test_create_dataset_fails_with_9_documents(self):
+        """Test boundary condition: 9 documents should fail."""
+        fn = _find_endpoint("create_dataset")
+
+        mock_db = MagicMock()
+
+        # 9 documents - just under the limit
+        request = DatasetCreateRequest(
+            name="test-dataset",
+            document_ids=TEST_DOC_UUIDS[:9],
+        )
+
+        from fastapi import HTTPException
+
+        with pytest.raises(HTTPException) as exc_info:
+            asyncio.run(fn(request=request, admin_token=TEST_TOKEN, db=mock_db))
+
+        assert exc_info.value.status_code == 400
+        assert "Minimum 10 documents required" in exc_info.value.detail
+
+    def test_create_dataset_succeeds_with_exactly_10_documents(self):
+        """Test boundary condition: exactly 10 documents should succeed."""
+        fn = _find_endpoint("create_dataset")
+
+        mock_db = MagicMock()
+        mock_db.create_dataset.return_value = _make_dataset(status="building")
+
+        mock_builder = MagicMock()
+
+        # Exactly 10 documents - should pass
+        request = DatasetCreateRequest(
+            name="test-dataset",
+            document_ids=TEST_DOC_UUIDS[:10],
+        )
+
+        with patch(
+            "inference.web.services.dataset_builder.DatasetBuilder",
+            return_value=mock_builder,
+        ):
+            result = asyncio.run(fn(request=request, admin_token=TEST_TOKEN, db=mock_db))
+
+        mock_db.create_dataset.assert_called_once()
+        assert result.dataset_id == TEST_DATASET_UUID
+
 
 class TestListDatasetsRoute:
     """Tests for GET /admin/training/datasets."""
@@ -198,3 +268,53 @@ class TestTrainFromDatasetRoute:
         with pytest.raises(HTTPException) as exc_info:
             asyncio.run(fn(dataset_id=TEST_DATASET_UUID, request=request, admin_token=TEST_TOKEN, db=mock_db))
         assert exc_info.value.status_code == 400
+
+    def test_incremental_training_with_base_model(self):
+        """Test training with base_model_version_id for incremental training."""
+        fn = _find_endpoint("train_from_dataset")
+
+        mock_model_version = MagicMock()
+        mock_model_version.model_path = "runs/train/invoice_fields/weights/best.pt"
+        mock_model_version.version = "1.0.0"
+
+        mock_db = MagicMock()
+        mock_db.get_dataset.return_value = _make_dataset(status="ready")
+        mock_db.get_model_version.return_value = mock_model_version
+        mock_db.create_training_task.return_value = TEST_TASK_UUID
+
+        base_model_uuid = "550e8400-e29b-41d4-a716-446655440099"
+        config = TrainingConfig(base_model_version_id=base_model_uuid)
+        request = DatasetTrainRequest(name="incremental-train", config=config)
+
+        result = asyncio.run(fn(dataset_id=TEST_DATASET_UUID, request=request, admin_token=TEST_TOKEN, db=mock_db))
+
+        # Verify model version was looked up
+        mock_db.get_model_version.assert_called_once_with(base_model_uuid)
+
+        # Verify task was created with finetune type
+        call_kwargs = mock_db.create_training_task.call_args[1]
+        assert call_kwargs["task_type"] == "finetune"
+        assert call_kwargs["config"]["base_model_path"] == "runs/train/invoice_fields/weights/best.pt"
+        assert call_kwargs["config"]["base_model_version"] == "1.0.0"
+
+        assert result.task_id == TEST_TASK_UUID
+        assert "Incremental training" in result.message
+
+    def test_incremental_training_with_invalid_base_model_fails(self):
+        """Test that training fails if base_model_version_id doesn't exist."""
+        fn = _find_endpoint("train_from_dataset")
+
+        mock_db = MagicMock()
+        mock_db.get_dataset.return_value = _make_dataset(status="ready")
+        mock_db.get_model_version.return_value = None
+
+        base_model_uuid = "550e8400-e29b-41d4-a716-446655440099"
+        config = TrainingConfig(base_model_version_id=base_model_uuid)
+        request = DatasetTrainRequest(name="incremental-train", config=config)
+
+        from fastapi import HTTPException
+
+        with pytest.raises(HTTPException) as exc_info:
+            asyncio.run(fn(dataset_id=TEST_DATASET_UUID, request=request, admin_token=TEST_TOKEN, db=mock_db))
+        assert exc_info.value.status_code == 404
+        assert "Base model version not found" in exc_info.value.detail
diff --git a/tests/web/test_model_versions.py b/tests/web/test_model_versions.py
new file mode 100644
index 0000000..e28249a
--- /dev/null
+++ b/tests/web/test_model_versions.py
@@ -0,0 +1,399 @@
+"""
+Tests for Model Version API routes.
+"""
+
+import asyncio
+from datetime import datetime, timezone
+from unittest.mock import MagicMock
+from uuid import UUID
+
+import pytest
+
+from inference.data.admin_models import ModelVersion
+from inference.web.api.v1.admin.training import create_training_router
+from inference.web.schemas.admin import (
+    ModelVersionCreateRequest,
+    ModelVersionUpdateRequest,
+)
+
+
+TEST_VERSION_UUID = "880e8400-e29b-41d4-a716-446655440020"
+TEST_VERSION_UUID_2 = "880e8400-e29b-41d4-a716-446655440021"
+TEST_TASK_UUID = "770e8400-e29b-41d4-a716-446655440002"
+TEST_DATASET_UUID = "880e8400-e29b-41d4-a716-446655440010"
+TEST_TOKEN = "test-admin-token-12345"
+
+
+def _make_model_version(**overrides) -> MagicMock:
+    """Create a mock ModelVersion."""
+    defaults = dict(
+        version_id=UUID(TEST_VERSION_UUID),
+        version="1.0.0",
+        name="test-model-v1",
+        description="Test model version",
+        model_path="/models/test-model-v1.pt",
+        status="inactive",
+        is_active=False,
+        task_id=UUID(TEST_TASK_UUID),
+        dataset_id=UUID(TEST_DATASET_UUID),
+        metrics_mAP=0.935,
+        metrics_precision=0.92,
+        metrics_recall=0.88,
+        document_count=100,
+        training_config={"epochs": 100, "batch_size": 16},
+        file_size=52428800,
+        trained_at=datetime(2025, 1, 15, tzinfo=timezone.utc),
+        activated_at=None,
+        created_at=datetime(2025, 1, 1, tzinfo=timezone.utc),
+        updated_at=datetime(2025, 1, 15, tzinfo=timezone.utc),
+    )
+    defaults.update(overrides)
+    model = MagicMock(spec=ModelVersion)
+    for k, v in defaults.items():
+        setattr(model, k, v)
+    return model
+
+
+def _find_endpoint(name: str):
+    """Find endpoint function by name."""
+    router = create_training_router()
+    for route in router.routes:
+        if hasattr(route, "endpoint") and route.endpoint.__name__ == name:
+            return route.endpoint
+    raise AssertionError(f"Endpoint {name} not found")
+
+
+class TestModelVersionRouterRegistration:
+    """Tests that model version endpoints are registered."""
+
+    def test_router_has_model_endpoints(self):
+        router = create_training_router()
+        paths = [route.path for route in router.routes]
+        assert any("models" in p for p in paths)
+
+    def test_has_create_model_version_endpoint(self):
+        endpoint = _find_endpoint("create_model_version")
+        assert endpoint is not None
+
+    def test_has_list_model_versions_endpoint(self):
+        endpoint = _find_endpoint("list_model_versions")
+        assert endpoint is not None
+
+    def test_has_get_active_model_endpoint(self):
+        endpoint = _find_endpoint("get_active_model")
+        assert endpoint is not None
+
+    def test_has_activate_model_version_endpoint(self):
+        endpoint = _find_endpoint("activate_model_version")
+        assert endpoint is not None
+
+
+class TestCreateModelVersionRoute:
+    """Tests for POST /admin/training/models."""
+
+    def test_create_model_version(self):
+        fn = _find_endpoint("create_model_version")
+
+        mock_db = MagicMock()
+        mock_db.create_model_version.return_value = _make_model_version()
+
+        request = ModelVersionCreateRequest(
+            version="1.0.0",
+            name="test-model-v1",
+            model_path="/models/test-model-v1.pt",
+            description="Test model",
+            metrics_mAP=0.935,
+            document_count=100,
+        )
+
+        result = asyncio.run(fn(request=request, admin_token=TEST_TOKEN, db=mock_db))
+
+        mock_db.create_model_version.assert_called_once()
+        assert result.version_id == TEST_VERSION_UUID
+        assert result.status == "inactive"
+        assert result.message == "Model version created successfully"
+
+    def test_create_model_version_with_task_and_dataset(self):
+        fn = _find_endpoint("create_model_version")
+
+        mock_db = MagicMock()
+        mock_db.create_model_version.return_value = _make_model_version()
+
+        request = ModelVersionCreateRequest(
+            version="1.0.0",
+            name="test-model-v1",
+            model_path="/models/test-model-v1.pt",
+            task_id=TEST_TASK_UUID,
+            dataset_id=TEST_DATASET_UUID,
+        )
+
+        result = asyncio.run(fn(request=request, admin_token=TEST_TOKEN, db=mock_db))
+
+        call_kwargs = mock_db.create_model_version.call_args[1]
+        assert call_kwargs["task_id"] == TEST_TASK_UUID
+        assert call_kwargs["dataset_id"] == TEST_DATASET_UUID
+
+
+class TestListModelVersionsRoute:
+    """Tests for GET /admin/training/models."""
+
+    def test_list_model_versions(self):
+        fn = _find_endpoint("list_model_versions")
+
+        mock_db = MagicMock()
+        mock_db.get_model_versions.return_value = (
+            [_make_model_version(), _make_model_version(version_id=UUID(TEST_VERSION_UUID_2), version="1.1.0")],
+            2,
+        )
+
+        result = asyncio.run(fn(admin_token=TEST_TOKEN, db=mock_db, status=None, limit=20, offset=0))
+
+        assert result.total == 2
+        assert len(result.models) == 2
+        assert result.models[0].version == "1.0.0"
+
+    def test_list_model_versions_with_status_filter(self):
+        fn = _find_endpoint("list_model_versions")
+
+        mock_db = MagicMock()
+        mock_db.get_model_versions.return_value = ([_make_model_version(status="active", is_active=True)], 1)
+
+        result = asyncio.run(fn(admin_token=TEST_TOKEN, db=mock_db, status="active", limit=20, offset=0))
+
+        mock_db.get_model_versions.assert_called_once_with(status="active", limit=20, offset=0)
+        assert result.total == 1
+        assert result.models[0].status == "active"
+
+
+class TestGetActiveModelRoute:
+    """Tests for GET /admin/training/models/active."""
+
+    def test_get_active_model_when_exists(self):
+        fn = _find_endpoint("get_active_model")
+
+        mock_db = MagicMock()
+        mock_db.get_active_model_version.return_value = _make_model_version(status="active", is_active=True)
+
+        result = asyncio.run(fn(admin_token=TEST_TOKEN, db=mock_db))
+
+        assert result.has_active_model is True
+        assert result.model is not None
+        assert result.model.is_active is True
+
+    def test_get_active_model_when_none(self):
+        fn = _find_endpoint("get_active_model")
+
+        mock_db = MagicMock()
+        mock_db.get_active_model_version.return_value = None
+
+        result = asyncio.run(fn(admin_token=TEST_TOKEN, db=mock_db))
+
+        assert result.has_active_model is False
+        assert result.model is None
+
+
+class TestGetModelVersionRoute:
+    """Tests for GET /admin/training/models/{version_id}."""
+
+    def test_get_model_version(self):
+        fn = _find_endpoint("get_model_version")
+
+        mock_db = MagicMock()
+        mock_db.get_model_version.return_value = _make_model_version()
+
+        result = asyncio.run(fn(version_id=TEST_VERSION_UUID, admin_token=TEST_TOKEN, db=mock_db))
+
+        assert result.version_id == TEST_VERSION_UUID
+        assert result.version == "1.0.0"
+        assert result.name == "test-model-v1"
+        assert result.metrics_mAP == 0.935
+
+    def test_get_model_version_not_found(self):
+        fn = _find_endpoint("get_model_version")
+
+        mock_db = MagicMock()
+        mock_db.get_model_version.return_value = None
+
+        from fastapi import HTTPException
+
+        with pytest.raises(HTTPException) as exc_info:
+            asyncio.run(fn(version_id=TEST_VERSION_UUID, admin_token=TEST_TOKEN, db=mock_db))
+        assert exc_info.value.status_code == 404
+
+
+class TestUpdateModelVersionRoute:
+    """Tests for PATCH /admin/training/models/{version_id}."""
+
+    def test_update_model_version(self):
+        fn = _find_endpoint("update_model_version")
+
+        mock_db = MagicMock()
+        mock_db.update_model_version.return_value = _make_model_version(name="updated-name")
+
+        request = ModelVersionUpdateRequest(name="updated-name", description="Updated description")
+
+        result = asyncio.run(fn(version_id=TEST_VERSION_UUID, request=request, admin_token=TEST_TOKEN, db=mock_db))
+
+        mock_db.update_model_version.assert_called_once_with(
+            version_id=TEST_VERSION_UUID,
+            name="updated-name",
+            description="Updated description",
+            status=None,
+        )
+        assert result.message == "Model version updated successfully"
+
+    def test_update_model_version_not_found(self):
+        fn = _find_endpoint("update_model_version")
+
+        mock_db = MagicMock()
+        mock_db.update_model_version.return_value = None
+
+        request = ModelVersionUpdateRequest(name="updated-name")
+
+        from fastapi import HTTPException
+
+        with pytest.raises(HTTPException) as exc_info:
+            asyncio.run(fn(version_id=TEST_VERSION_UUID, request=request, admin_token=TEST_TOKEN, db=mock_db))
+        assert exc_info.value.status_code == 404
+
+
+class TestActivateModelVersionRoute:
+    """Tests for POST /admin/training/models/{version_id}/activate."""
+
+    def test_activate_model_version(self):
+        fn = _find_endpoint("activate_model_version")
+
+        mock_db = MagicMock()
+        mock_db.activate_model_version.return_value = _make_model_version(status="active", is_active=True)
+
+        result = asyncio.run(fn(version_id=TEST_VERSION_UUID, admin_token=TEST_TOKEN, db=mock_db))
+
+        mock_db.activate_model_version.assert_called_once_with(TEST_VERSION_UUID)
+        assert result.status == "active"
+        assert result.message == "Model version activated for inference"
+
+    def test_activate_model_version_not_found(self):
+        fn = _find_endpoint("activate_model_version")
+
+        mock_db = MagicMock()
+        mock_db.activate_model_version.return_value = None
+
+        from fastapi import HTTPException
+
+        with pytest.raises(HTTPException) as exc_info:
+            asyncio.run(fn(version_id=TEST_VERSION_UUID, admin_token=TEST_TOKEN, db=mock_db))
+        assert exc_info.value.status_code == 404
+
+
+class TestDeactivateModelVersionRoute:
+    """Tests for POST /admin/training/models/{version_id}/deactivate."""
+
+    def test_deactivate_model_version(self):
+        fn = _find_endpoint("deactivate_model_version")
+
+        mock_db = MagicMock()
+        mock_db.deactivate_model_version.return_value = _make_model_version(status="inactive", is_active=False)
+
+        result = asyncio.run(fn(version_id=TEST_VERSION_UUID, admin_token=TEST_TOKEN, db=mock_db))
+
+        assert result.status == "inactive"
+        assert result.message == "Model version deactivated"
+
+    def test_deactivate_model_version_not_found(self):
+        fn = _find_endpoint("deactivate_model_version")
+
+        mock_db = MagicMock()
+        mock_db.deactivate_model_version.return_value = None
+
+        from fastapi import HTTPException
+
+        with pytest.raises(HTTPException) as exc_info:
+            asyncio.run(fn(version_id=TEST_VERSION_UUID, admin_token=TEST_TOKEN, db=mock_db))
+        assert exc_info.value.status_code == 404
+
+
+class TestArchiveModelVersionRoute:
+    """Tests for POST /admin/training/models/{version_id}/archive."""
+
+    def test_archive_model_version(self):
+        fn = _find_endpoint("archive_model_version")
+
+        mock_db = MagicMock()
+        mock_db.archive_model_version.return_value = _make_model_version(status="archived")
+
+        result = asyncio.run(fn(version_id=TEST_VERSION_UUID, admin_token=TEST_TOKEN, db=mock_db))
+
+        assert result.status == "archived"
+        assert result.message == "Model version archived"
+
+    def test_archive_active_model_fails(self):
+        fn = _find_endpoint("archive_model_version")
+
+        mock_db = MagicMock()
+        mock_db.archive_model_version.return_value = None
+
+        from fastapi import HTTPException
+
+        with pytest.raises(HTTPException) as exc_info:
+            asyncio.run(fn(version_id=TEST_VERSION_UUID, admin_token=TEST_TOKEN, db=mock_db))
+        assert exc_info.value.status_code == 400
+
+
+class TestDeleteModelVersionRoute:
+    """Tests for DELETE /admin/training/models/{version_id}."""
+
+    def test_delete_model_version(self):
+        fn = _find_endpoint("delete_model_version")
+
+        mock_db = MagicMock()
+        mock_db.delete_model_version.return_value = True
+
+        result = asyncio.run(fn(version_id=TEST_VERSION_UUID, admin_token=TEST_TOKEN, db=mock_db))
+
+        mock_db.delete_model_version.assert_called_once_with(TEST_VERSION_UUID)
+        assert result["message"] == "Model version deleted"
+
+    def test_delete_active_model_fails(self):
+        fn = _find_endpoint("delete_model_version")
+
+        mock_db = MagicMock()
+        mock_db.delete_model_version.return_value = False
+
+        from fastapi import HTTPException
+
+        with pytest.raises(HTTPException) as exc_info:
+            asyncio.run(fn(version_id=TEST_VERSION_UUID, admin_token=TEST_TOKEN, db=mock_db))
+        assert exc_info.value.status_code == 400
+
+
+class TestModelVersionSchemas:
+    """Tests for model version Pydantic schemas."""
+
+    def test_create_request_validation(self):
+        request = ModelVersionCreateRequest(
+            version="1.0.0",
+            name="test-model",
+            model_path="/models/test.pt",
+        )
+        assert request.version == "1.0.0"
+        assert request.name == "test-model"
+        assert request.document_count == 0
+
+    def test_create_request_with_metrics(self):
+        request = ModelVersionCreateRequest(
+            version="2.0.0",
+            name="test-model-v2",
+            model_path="/models/v2.pt",
+            metrics_mAP=0.95,
+            metrics_precision=0.92,
+            metrics_recall=0.88,
+            document_count=500,
+        )
+        assert request.metrics_mAP == 0.95
+        assert request.document_count == 500
+
+    def test_update_request_partial(self):
+        request = ModelVersionUpdateRequest(name="new-name")
+        assert request.name == "new-name"
+        assert request.description is None
+        assert request.status is None