An automobile company collects several terabytes of sensor data from vehicles every day and stores...

SageMaker AI batch transform is the best option because the company runs inference only once each week on large datasets stored in Amazon S3. Batch transform is designed for offline batch inference, so the company does not need to deploy or maintain a persistent real-time endpoint. The same managed SageMaker workflow can process input records, run the model, and write inference results back to S3 with minimal infrastructure management. AWS Batch and Amazon EMR can process large datasets, but they introduce extra services and operational responsibility. AWS Glue is useful for ETL, but the option still incorrectly uses a SageMaker processing job for the inference workload rather than the managed batch inference feature. Therefore, batch transform provides the least operational overhead.