AWS Certified AI Practitioner Practice Exam

In Amazon SageMaker, which option is designed for handling model inference requests that can be delayed?

1. Real-time inference

2. Batch transformation

3. Serverless inference

4. Asynchronous inference

The correct answer is: Asynchronous inference

The correct choice is designed to handle model inference requests that can be delayed because it allows for processing requests without requiring immediate results. Asynchronous inference is particularly useful when dealing with large datasets or complex models where the processing time may exceed acceptable limit for real-time applications. Users can submit inference requests and receive the results at a later time, enabling them to continue with other tasks without waiting. In contrast, real-time inference requires immediate responses to requests, which is suitable for applications needing quick outputs. Batch transformation, while efficient for processing large volumes of data, is primarily focused on running inferences on batch data rather than individual requests. Serverless inference, while also designed to simplify deployment and scaling, is closely aligned with the real-time and on-demand invocation model rather than handling delayed requests. Thus, for scenarios where the response timing is flexible, asynchronous inference is the ideal choice.