Monitoring Types
- Data quality monitoring compares incoming data to a baseline schema and distribution.
- Model quality monitoring compares predictions to ground truth labels when available.
- Bias drift checks whether bias metrics change over time.
- Explainability drift checks whether feature attribution changes over time.
Operational Pattern
- Create a baseline from representative data.
- Capture production inference data.
- Schedule monitoring jobs.
- Send violations to CloudWatch and trigger review or retraining.
Practice Model Monitor Questions
Put your knowledge to the test with practice questions.