AWS Certified AI Practitioner Practice Exam

What inference option should a company use when analyzing large archived datasets in Amazon SageMaker?

1. Batch transformation

2. Real-time inference

3. Serverless inference

4. Asynchronous inference

The correct answer is: Batch transformation

When a company is analyzing large archived datasets in Amazon SageMaker, the most suitable inference option is batch transformation. This method allows for the processing of large volumes of data all at once rather than requiring real-time responses. Batch transformation is designed specifically for scenarios where predictions are needed for multiple records simultaneously, making it efficient for handling large datasets that may not require immediate results. In this case, archived datasets can be processed in batches, which optimizes resource usage and can reduce the overall cost of running inference jobs, as these jobs can be executed without the need for continuous provisioning of compute resources. Choosing batch transformation allows the company to take advantage of the scale and cost-effectiveness of processing large datasets without committing to a continuous, real-time infrastructure. This is particularly important when working with historical data that may be analyzed periodically rather than in real-time. The other inference options are better suited for different contexts; for example, real-time inference is aimed at applications requiring immediate predictions, while serverless and asynchronous inference cater to different resource management and latency needs respectively.