AI

If the detected AI workload is LLM-based (generative models), load the source-specific design reference instead of this file:

• If ai-workload-profile.json → summary.ai_source = "gemini": load references/design-refs/ai-gemini-to-bedrock.md
• If ai-workload-profile.json → summary.ai_source = "openai": load references/vendored/ai/ai-openai-to-bedrock.md
• If ai-workload-profile.json → summary.ai_source = "both": load both files
• If ai-workload-profile.json → summary.ai_source = "other" or absent, OR if the workload is traditional ML (custom models, Vision API, Speech API): use the SageMaker/Rekognition/Textract rubric below.

Signals (Decision Criteria)

Vertex AI (Endpoints / Models)

• Custom model inference → SageMaker Endpoints
• Pre-built model APIs → AWS APIs (Rekognition, Textract, Translate, etc.)
• Batch prediction → SageMaker Batch Transform

Cloud Vision API

• Image classification, OCR → AWS Rekognition (images) or Textract (OCR)
• Document understanding → AWS Textract (more powerful for docs)

Cloud Speech-to-Text

• Audio/video transcription (batch or streaming) → AWS Transcribe
• Call center audio (with sentiment/topic detection needs) → AWS Transcribe Call Analytics

Document AI

• Form/invoice/receipt/ID extraction → AWS Textract (AnalyzeExpense, AnalyzeID, or AnalyzeDocument with FORMS/TABLES depending on document type)
• General document OCR → AWS Textract (DetectDocumentText)

Cloud ML Engine (deprecated — legacy Terraform configs only)

• Model training → SageMaker (managed training jobs)
• AutoML → SageMaker Autopilot / Canvas

6-Criteria Rubric

Apply in order:

1. Eliminators: Does GCP config require AWS-unsupported features? If yes: use alternative
2. Operational Model: Managed (SageMaker) vs Custom (EC2 + training)?
   • Prefer managed
3. User Preference: From preferences.json: design_constraints.cost_sensitivity + ai_constraints (if present)
   • If cost-sensitive → check SageMaker Spot + Autopilot
4. Feature Parity: Does GCP config need model type unavailable in AWS?
   • Example: TensorFlow 2.x → SageMaker (supported)
5. Cluster Context: Are other compute resources running ML? Prefer SageMaker affinity
6. Simplicity: SageMaker endpoints (managed) > custom EC2 instances

Examples

Example 1: Vertex AI Endpoint (PyTorch model)

• GCP: google_ai_platform_model (model_name=”image-classifier”, framework=PYTORCH)
• Signals: Custom model inference, PyTorch
• Criterion 1 (Eliminators): PASS (PyTorch supported)
• Criterion 2 (Operational Model): SageMaker Endpoint (managed)
• → AWS: SageMaker Endpoint (PyTorch container)
• Confidence: inferred

Example 2: Cloud Vision API

• GCP: google_vision_api_call (feature=TEXT_DETECTION, image_source=GCS)
• Signals: Pre-built API
• → AWS: Textract (if document OCR) or Rekognition (if image classification)
• Confidence: inferred

Example 3: AutoML (image classification)

• GCP: google_automl_image_classification_dataset
• Signals: Training pipeline, classification
• Criterion 1 (Eliminators): PASS
• Criterion 2 (Operational Model): SageMaker Autopilot (managed)
• → AWS: SageMaker Autopilot + Canvas (for low-code)
• Confidence: inferred

Example 4: Document AI (workload capability document_extraction)

• GCP: documentai.process_document call detected in app code
• Signals: Pre-built document extraction API, no custom model training
• → AWS: Textract (AnalyzeExpense for invoices/receipts, AnalyzeID for identity documents, AnalyzeDocument with FORMS/TABLES for general forms, DetectDocumentText for plain OCR — pick the closest match to the detected GCP processor type; default to AnalyzeDocument with FORMS/TABLES when the processor type can’t be determined)
• Confidence: inferred

Example 5: Cloud Speech-to-Text (workload capability speech_transcription)

• GCP: speech.recognize or speech.long_running_recognize call detected in app code
• Signals: Pre-built speech-to-text API
• → AWS: Transcribe (streaming → StartStreamTranscription; batch → StartTranscriptionJob)
• Confidence: inferred

Output Schema