For an ML model in production, a company wants to work on monitoring performance degradation after deployment and on documenting the model's intended use and risks for a governance audit. Which TWO AWS features help with these? (Choose TWO.)
Select two AWS features that help with monitoring and documenting a model.
SageMaker Clarify
SageMaker Clarify is a feature that provides bias detection and prediction explainability.
It is used in the same responsible AI context, but it does not address this question's requirements (performance degradation monitoring and documentation), so it is incorrect.
SageMaker Model Monitor
This is correct. SageMaker Model Monitor is a feature that monitors data/quality drift (performance degradation) of a model after deployment.
Amazon Augmented AI (A2I)
Amazon A2I is a service that incorporates human review of AI predictions.
It is a means of ensuring quality, but it is not a feature for continuous performance degradation monitoring or for documenting intended use and risks, so it is incorrect.
SageMaker JumpStart
SageMaker JumpStart is a model hub for getting started right away with pretrained models and ready-made solutions.
It is a feature that helps obtain models, not one that performs post-operation monitoring or documentation, so it is incorrect.
SageMaker Model Cards
This is correct. SageMaker Model Cards is a feature that documents a model's intended use, performance, known risks, and considerations, and supports governance.
Distinguish SageMaker governance features by their roles.
- Model Monitor: post-deployment drift and performance degradation monitoring.
- Model Cards: documentation of use, performance, and risks (for audits).
- Clarify: bias detection and explainability.
- A2I: incorporating human review.
- JumpStart: hub of pretrained models.
All belong to the SageMaker family and are easily confused, so match the requirement's verb (monitor / document / detect / review).