Question

A lending company deployed a credit-approval model to a real-time SageMaker endpoint. The compliance team requires ongoing verification that the model does not develop disparate approval rates across applicant age groups after deployment, and wants automated alerts if fairness metrics degrade over time compared to a baseline established at training. Which approach best meets this requirement with the least custom development?

Accepted Answer

Enable SageMaker Model Monitor's bias drift monitoring with SageMaker Clarify, configure a bias baseline from training data, and schedule monitoring jobs that emit CloudWatch metrics for alarms.

Answer

Enable SageMaker Model Monitor's data quality monitoring to detect statistical changes in the age feature distribution and alarm on any deviation.

Answer

Write a scheduled Lambda function that pulls captured inference data from S3, computes approval rates per age group with pandas, and publishes custom CloudWatch metrics.

Answer

Enable SageMaker Model Monitor's model quality monitoring with ground-truth labels to track accuracy degradation across the full population.