Machine Learning
Amazon SageMaker Model Registry
Amazon SageMaker Model Registry catalogs model packages, versions, metadata, metrics, and approval status. It creates a controlled handoff from experimentation to deployment for custom models and inference artifacts.
Key points
- Model groups organize related versions
- Approval status can gate automated pipelines
- Packages can reference inference images, model data, and validation profiles
When to use it
- Approve a reranker only after safety and latency tests
- Trace a deployed endpoint to immutable model and container artifacts
Exam tips
- Use Model Registry for model lifecycle evidence, not for prompt versioning
- Approval should reference reproducible datasets, images, thresholds, and intended use rather than a score alone