Automate SageMaker Retraining with S3 Events and Lambda

Quick Answer

This question tests whether you can map an event-driven automation pattern onto AWS's serverless building blocks rather than reaching for compute you'd have to manage yourself. Amazon S3 is the natural trigger point because it can emit object-creation events the moment new training data lands in the bucket, without any polling or scheduled job. AWS Lambda is the natural glue between that event and the training system: it runs briefly in response to the S3 notification, and its only job is to call the SageMaker API to kick off a new training job, without needing to run continuously or manage its own infrastructure. Amazon SageMaker is the service that actually owns the retraining work itself, spinning up managed training infrastructure and producing an updated model artifact. Together these three services form a clean trigger-orchestrate-execute chain with no persistent servers to patch or scale. EC2 is the wrong choice here because it would mean provisioning and managing your own compute for a task SageMaker already handles as a managed service, adding operational overhead with no benefit. AWS Glue is designed for data transformation and ETL, not for triggering or running model training, so it doesn't fit this workflow. When you see a scenario asking for an automated, event-driven pipeline reacting to new data in S3, expect the answer to combine an S3 event source with a lightweight orchestrator like Lambda and the managed service that performs the actual work.