AWS Certified Machine Learning – SpecialtyMachine Learning Implementation and OperationsHard
A healthcare provider is deploying a diagnostic ML model that processes sensitive patient data. Regulatory compliance (e.g., HIPAA) requires strict data provenance and an immutable audit trail of every inference request, including the input data, model version, and prediction results. This audit trail is essential for post-hoc analysis, debugging, and demonstrating compliance. Which SageMaker feature, combined with appropriate AWS logging, can best meet these requirements?
- ASageMaker Pipelines with model versioning.
- BSageMaker Asynchronous Inference with data capture.
- CSageMaker Real-Time Endpoint with data capture enabled.
- DSageMaker Model Monitor with custom metrics.
Show answer & explanationAnswer & explanation
Correct answer: C. SageMaker Real-Time Endpoint with data capture enabled.
SageMaker Real-Time Endpoints can be configured with data capture, which automatically saves a configurable percentage of inference requests and responses (including input data and prediction results) to an S3 bucket. This raw data, combined with CloudWatch logs for model versioning and timestamps, creates an immutable audit trail necessary for regulatory compliance and deep analysis.
Why the other options are wrong
- A. SageMaker Pipelines manage model versioning and lineage during development and deployment, but they don't capture every live inference request and response at the endpoint for auditing purposes.
- B. Asynchronous Inference captures data, but it's for non-real-time scenarios. The question implies a diagnostic model, which often requires real-time predictions. Real-time endpoints also support data capture.
- D. Model Monitor analyzes captured data for drift but doesn't itself create the detailed, raw audit trail of every request and response.
SageMaker Data Capture
A feature for SageMaker inference endpoints (real-time or asynchronous) that automatically captures and saves input and output data for inference requests to S3.
- Provides a raw, immutable record of inferences
- Essential for auditing, debugging, and compliance
- Captures request payload, response payload, and metadata
- Configurable sampling rate for capture
Memory trick: Capture every prediction, keep the audit trail clear.