An ML pipeline should automatically start a retraining job whenever data drift is detected. How should the engineer configure the pipeline to detect drift and trigger retraining?

SageMaker Model Monitor is specifically designed to detect data and model quality drift in deployed machine learning models. It can continuously analyze data flowing into endpoints and compare it against a baseline, generating metrics and alerts when drift is detected. These alerts can then trigger an AWS Lambda function, which can orchestrate the retraining pipeline. AWS Glue crawlers and ETL jobs are primarily for data cataloging and transformation, not real-time drift detection for ML models. Amazon Managed Service for Apache Flink is for real-time stream processing, but SageMaker Model Monitor offers a more integrated and specialized solution for ML drift. Amazon QuickSight is a business intelligence service for visualization and anomaly detection in dashboards, not typically used for triggering ML retraining workflows directly.