Prepare for the AWS Certified Machine Learning Engineer - Associate exam with 500+ real exam-style practice questions. Our question bank covers all four domains: data preparation for ML, model development, deployment and orchestration of ML workflows, and ML solution monitoring, maintenance, and security.
Our 500+ practice questions are distributed across these domains to match the real exam weighting.
The AWS Certified Machine Learning Engineer - Associate certification validates your ability to build, deploy, and maintain ML solutions and pipelines on AWS, with Amazon SageMaker at the center of nearly every exam domain.
This exam goes beyond theory: it tests hands-on decisions like choosing between real-time, serverless, asynchronous, and batch inference; setting up CI/CD for model pipelines; detecting data and concept drift with Model Monitor; and securing ML systems with IAM and VPC isolation.
Candidates consistently report that endpoint selection, evaluation-metric tradeoffs, and ML CI/CD scenarios are the hardest parts of the exam. Structured practice with scenario-based questions is the fastest way to build the judgment these questions demand.
Try each question before revealing the answer.
A team trains a computer-vision model on millions of small image files in S3. Training jobs wait 40+ minutes at startup while the dataset downloads. The training script reads files via normal POSIX calls and must not change. Which SageMaker input mode should they use?
Fast File mode exposes S3 objects as a POSIX file system and streams them on demand, so jobs start almost immediately and the existing file-based training script works unchanged.
A fraud-detection model must catch as many fraudulent transactions as possible, even at the cost of some false alarms. Which evaluation metric should be prioritized?
Recall measures the fraction of actual positives the model catches. When missing a fraud case is costlier than a false alarm, recall is the metric to maximize — typically by lowering the classification threshold.
An ML team wants production traffic gradually shifted to a new model version, with automatic rollback if CloudWatch alarms fire during the shift. Which deployment strategy fits?
Blue/green deployment with canary traffic shifting sends a small slice of traffic to the new version first and monitors CloudWatch alarms; if alarms trigger, traffic automatically rolls back to the old version.
Start practicing with 500+ exam-style questions. 100% free during preview — no credit card required.