How Should SageMaker AI Serve High-Resolution Image Cost Predictions with User Notifications?
A company is developing an internal cost-estimation tool that uses an ML model in Amazon SageMaker AI. Users upload high-resolution images to the tool. The model must process each image and predict the cost of the object in the image. The model also must notify the user when processing is complete. Which solution will meet these requirements?
Community Insight
The question tests matching a long-running, large-payload inference workload to SageMaker asynchronous inference plus push notification, and the trap is assuming batch transform or EFS storage can deliver per-request user alerts.
This scenario tests how to pick the right SageMaker AI inference pattern and notification service for large image payloads that take time to process. The page establishes that asynchronous inference backed by Amazon S3 storage and an Amazon SNS topic is the design that satisfies both the inference and the completion-notification requirements.
Many candidates pick a batch transform option because batch processing sounds right for images, but batch transform has no built-in per-request completion callback for an end user and it cannot serve interactive notifications.
Community Discussion (2 comments)
Comments & Corrections
No comments yet — spotted an error or have a note? Share it below.
Expert Analysis
Why the Answer Is Correct
High-resolution images produce large payloads and take a relatively long time to process, which is exactly the profile SageMaker AI asynchronous inference is built for: the request is queued, the endpoint returns immediately, and the result is written to Amazon S3 when the job finishes. Amazon S3 is the natural ingestion point because users can upload directly to a bucket and the endpoint can read the object with the invocation. The remaining requirement — notifying the user when processing is complete — is a push-style fan-out, and an Amazon SNS topic is the managed pub/sub service designed for that, delivering email, SMS, or webhook notifications to subscribers. Combining S3 storage, asynchronous inference, and SNS covers storage, long-running inference, and notification in one coherent design, which is why option B satisfies every stated requirement.Why the Other Options Are Wrong
Options A and C both rely on batch transform, which is designed for scheduled bulk scoring of an entire dataset rather than on-demand per-image prediction with a completion callback; batch transform also requires the input data to already be in S3 and offers no native notification channel to the requesting user. Options C and D store images on Amazon EFS, which is a shared POSIX file system for compute instances, not the object store that SageMaker inference and transform jobs consume input from, so it adds a needless and mismatched storage layer. Option A's Amazon SQS queue is a pull-based mechanism: it holds messages until a consumer polls them, so it does not notify the end user that their image has been processed. Option D pairs the correct asynchronous inference mode with the wrong storage service, so it fails the storage requirement even though the notification half is sound.Community Comment Notes
Both community respondents converge on option B, with one noting that "S3 because images can't be stored in EFS" — capturing the storage mismatch that eliminates C and D. The same commenter links asynchronous inference to the requirement that "the model must run and notify the user of the output", and pairs SNS with notifications. A second commenter simply summarizes the same reasoning as "S3, Asyn and SNS". The consensus among learners therefore aligns with the vendor-documented pattern of S3 input, asynchronous inference, and SNS completion alerts.Official Reference
Exam Strategy
When a stem mentions large payloads, long processing times, near-real-time per-request results, or a user-facing completion alert, map it to SageMaker asynchronous inference rather than batch transform. Also default to Amazon S3 as the input store for SageMaker inference unless the stem explicitly describes a shared file system mounted on a training or processing instance.
Frequently Asked Questions
Why is batch transform wrong for notifying users after each image is processed?
Batch transform scores a whole dataset on a schedule and writes results to S3; it has no per-request callback, so users would not be alerted when their individual image finishes.
Why not store the uploaded images in Amazon EFS instead of Amazon S3?
SageMaker inference reads input objects from Amazon S3, and EFS is a shared file system for compute instances rather than the object store used for inference payloads.
Related Analysis
Practice All MLA-C01 Questions
Access 115 questions with complete answers and detailed explanations.
View Full MLA-C01 Practice Test →