An ecommerce company wants to update a production real-time machine learning (ML) recommendation engine API that uses Amazon SageMaker. The company wants to release a new model but does not want to…

Creating a new SageMaker endpoint for the new model would require applications relying on the API to update their endpoint URL, violating the requirement to avoid changes to applications.

Modifying the existing endpoint to use SageMaker production variants allows hosting multiple models behind a single endpoint, enabling A/B testing or canary deployments by distributing traffic between the old and new models without requiring changes to client applications, all with minimal operational overhead as SageMaker manages the infrastructure.

Amazon SageMaker batch transform is designed for offline, asynchronous inference on large datasets, not for real-time, low-latency API predictions needed for a recommendation engine.

Creating a new SageMaker endpoint for the new model, similar to option A, would require changes to dependent applications, failing to meet the specified requirement.