DataStream Inc. wants a repeatable CI/CD pipeline for model development that: triggers on Git commits, runs unit tests and static checks, triggers SageMaker Pipelines training, registers a model to the Model Registry, and promotes to a production endpoint after manual approval. The team prefers to use managed SageMaker tooling to provision the CI/CD infra (CodePipeline, CodeBuild, IAM roles). Which approach best meets their requirement with minimal hand-built pipelines and integrates with SageMaker Studio for developer onboarding?

The built-in SageMaker MLOps (CI/CD) project template directly addresses all requirements. It automatically provisions a complete CI/CD pipeline using AWS CodePipeline, CodeBuild, and SageMaker Pipelines, triggered by Git commits. This template includes stages for unit tests, static checks, model training via SageMaker Pipelines, model registration to the Model Registry, and manual approval for production deployment. This approach minimizes hand-built infrastructure and integrates seamlessly with SageMaker Studio for developer onboarding. Manually authoring CloudFormation templates is possible but requires significant effort to replicate the template's functionality. Using only CodeBuild lacks the full orchestration capabilities of CodePipeline, especially for multi-stage processes and manual approvals. A Lambda-backed custom webhook is overly complex and bypasses managed services like CodePipeline and SageMaker Pipelines, increasing operational overhead and reducing integration with SageMaker Studio.