A company's ML engineer has deployed an ML model for sentiment analysis to an Amazon SageMaker endpoint. The ML engineer needs to explain to company stakeholders how the model makes predictions…

Question

Options

• AUse SageMaker Model Monitor on the deployed model.
• BUse SageMaker Clarify on the deployed model.
• CShow the distribution of inferences from A/?testing in Amazon CloudWatch.
• DAdd a shadow endpoint. Analyze prediction differences on samples.

• A
  3% (1)
• B
  88% (29)
• C
  3% (1)
• D
  6% (2)

Explanation

SageMaker Clarify is purpose-built for model explainability - it uses SHAP (SHapley Additive exPlanations) values to show how each feature contributes to individual predictions, making it the right tool when stakeholders need to understand why a model made a specific decision.

Why the distractors are wrong:

• A (Model Monitor) tracks data drift and model quality over time - it detects when something goes wrong, not why the model predicts what it does.
• C (CloudWatch distributions) shows operational metrics like inference counts and latency, not feature-level prediction reasoning.
• D (Shadow endpoint) compares two model versions against each other for performance validation - it reveals differences between models, not explanations of predictions.

Memory tip: Think "Clarify = Clarity" - when a stakeholder asks "why did the model predict this?", you need clarity into the model's reasoning, which is exactly what SageMaker Clarify provides through feature attribution reports.

No community discussion yet for this question.