AWS Machine Learning Engineer Associate (MLA-C01) · 22% practice allocation

AWS MLA-C01: Deployment and Orchestration of ML Workflows Practice Questions

5 free sample questions from the Deployment and Orchestration of ML Workflows domain, with answers and explanations. The full MLA-C01 bank has 59 questions across 4 domains. Questions are scenario-based and written from the official AWS exam guide.

1.

A team must score a large S3 dataset once, with no need for a persistent endpoint, minimizing cost. Which SageMaker inference option fits BEST?

  • A. Batch transform✓ Correct
  • B. A 24/7 real-time endpoint
  • C. Serverless inference for each record separately
  • D. Manual predictions in a notebook forever
Explanation

Batch transform runs inference over a whole dataset without a persistent endpoint and shuts down after, minimizing cost for one-off/offline scoring. A 24/7 endpoint (B) wastes money when not needed continuously, and options C and D are inefficient for bulk scoring.

2.

A model must handle requests with large payloads and long processing times without blocking the client, using a queue-based pattern. Which SageMaker inference option is designed for this?

  • A. Asynchronous inference✓ Correct
  • B. Real-time endpoint with a 10 ms timeout
  • C. Batch transform on a schedule only
  • D. Serverless inference with no queue
Explanation

SageMaker Asynchronous Inference queues requests, handles large payloads and longer processing, and returns results without keeping the client blocked. A tight-timeout real-time endpoint (B) would fail on long jobs, scheduled batch (C) isn't request-driven, and serverless (D) targets intermittent small requests.

3.

An application needs low-latency, always-on predictions for each incoming user request. Which SageMaker inference option is MOST appropriate?

  • A. Real-time endpoint✓ Correct
  • B. Batch transform
  • C. Asynchronous inference for large payloads only
  • D. No deployment; run training jobs per request
Explanation

A SageMaker real-time endpoint provides persistent, low-latency inference for per-request predictions. Batch transform (B) processes datasets offline, asynchronous inference (C) suits large payloads/long processing with queuing, and running training jobs per request (D) is not an inference pattern.

4.

A team wants CI/CD for ML — automatically building, testing, and deploying models through a repeatable, governed process. Which SageMaker capability provides templates for this MLOps setup?

  • A. Amazon SageMaker Ground Truth
  • B. AWS Direct Connect
  • C. Amazon SageMaker Projects✓ Correct
  • D. Amazon Macie
Explanation

SageMaker Projects provides MLOps templates that set up CI/CD pipelines (with source/build/deploy stages) for automated, governed model delivery. Ground Truth (A) labels data, Macie (D) finds sensitive data, and Direct Connect (B) is networking — none provides ML CI/CD scaffolding.

5.

A model has sporadic, unpredictable traffic with idle periods, and the team wants to avoid paying for an always-on endpoint while still getting on-demand predictions. Which option fits?

  • A. SageMaker Serverless Inference✓ Correct
  • B. A provisioned real-time endpoint running 24/7
  • C. Batch transform triggered every second
  • D. Training a new model per request
Explanation

SageMaker Serverless Inference automatically provisions and scales compute (including to zero) for intermittent traffic, so you pay only for what you use. A 24/7 endpoint (B) pays during idle time, per-second batch transform (C) is impractical, and per-request training (D) is not inference.

Test yourself across every MLA-C01 domain

The free MLA-C01 practice test mixes about 20 questions weighted like the exam. The full 59-question bank adds instant grading, flashcards and a timed mock exam.

Take the free MLA-C01 practice testPractice the full bank

Study this domain

Deployment and Orchestration of ML Workflows study guide (MLA-C01) →

Other MLA-C01 domains

← All AWS Machine Learning Engineer Associate study resources