AWS Certified AI Practitioner
Detecting and Monitoring Bias
Putting numbers on fairness with SageMaker Clarify, watching for drift with Model Monitor, and routing uncertain predictions to people with Amazon A2I.
Intermediate 24 minutes 5 Learning Objectives
- Apply label quality analysis, human audits, and subgroup analysis to find bias without vendor tooling
- Compute and interpret Clarify pre-training and post-training bias metrics on a worked example
- Explain how SageMaker Model Monitor detects bias drift and feature attribution drift after deployment
- Describe when Amazon A2I routes a prediction to a human reviewer
- Select the right AWS tool for a described bias or quality problem