ML platform
SageMaker Clarify
Bias detection and explainability for data and model predictions.
Key points
- Helps detect potential bias in datasets and model predictions.
- Provides explainability reports so teams can understand feature impact on predictions.
- Can be used before training and after model deployment.
- Supports responsible AI practices by making model behavior easier to inspect.
- Complements monitoring and governance but does not fix bias automatically.
When to use it
- Choose Clarify when the question asks about bias detection or model explainability.
- Use it to compare outcomes across groups or inspect why a model made predictions.
- Use it during model review before approving a model for production.
Exam tips
- Clarify is for bias and explainability; Model Monitor is for drift and quality monitoring.
- Clarify does not replace human governance or domain review.
- Use it for custom ML in SageMaker, not for Bedrock Guardrails safety filtering.
- Bias can exist in data, labels, features, or model outputs.