AWS MLA-C01: Machine Learning Engineer Associate Practice Exam

What the AWS MLA-C01 exam covers

• Data Preparation for ML59 questions
• ML Model Development67 questions
• Deployment and Orchestration78 questions
• Monitoring and Security97 questions

Free AWS MLA-C01 sample questions

A sample of 10 questions with answers and explanations. Sign up free to practice all 301.

1. Question 1Data Preparation for ML

   Which AWS service is commonly used as the data lake for storing raw ML training data?

   • AAmazon SQS
   • BAWS KMS
   • CAmazon S3Correct
   • DAmazon Route 53
   ✓ Correct answer: C

   Amazon S3 is the primary and most cost-effective service for storing large-scale raw ML training data in a data lake architecture. S3 provides virtually unlimited storage, high durability, integration with SageMaker for training workflows, and excellent performance when used with services like Athena and Glue for data transformation.

   Why the other options are wrong

   • AAmazon SQS is a message queue service for event processing, not data storage.
   • BAWS KMS is a key management service for encryption, not data storage.
   • DAmazon Route 53 is a DNS service with no relevance to ML data lakes.
2. Question 2Monitoring and Security

   To launch a SageMaker training job inside a private VPC with no internet access, which create-training-job parameter specifies the subnets and security groups?

   • A--private-link-config '{"Subnets":["subnet-abc"]}'
   • B--vpc '{"Subnet":"subnet-abc","SG":"sg-123"}'
   • C--vpc-config '{"Subnets":["subnet-abc"],"SecurityGroupIds":["sg-123"]}'Correct
   • D--network-config '{"SubnetIds":["subnet-abc"]}'
   ✓ Correct answer: C

   Running in private VPC restricts data through security groups.

   Why the other options are wrong

   • APrivate subnets and VPC endpoints keep communication in AWS backbone. This architecture meets residency and compliance requirements. --private-link-config '{"Subnets":["subnet-abc"]}' is incorrect because this option does not satisfy the requirement.
   • B--vpc '{"Subnet":"subnet-abc","SG":"sg-123"}' is incorrect because this option does not satisfy the requirement.
   • D--network-config '{"SubnetIds":["subnet-abc"]}' is incorrect because this option does not satisfy the requirement.
3. Question 3Monitoring and Security

   A team wants to capture every inference request and response payload sent to a SageMaker real-time endpoint so they can later analyze input/output distributions. Which feature should they enable?

   • AVPC Flow Logs on the endpoint's subnet
   • BS3 server access logging on the model artifact bucket
   • CCloudFront access logs on the endpoint
   • DsageMaker Endpoint Data Capture, which writes request/response payloads to S3Correct
   ✓ Correct answer: D

   This correctly addresses the requirement following AWS best practices.

   Why the other options are wrong

   • AVPC Flow Logs on the endpoint's subnet is incorrect because this option does not satisfy the requirement.
   • BS3 server access logging on the model artifact bucket is incorrect because this option does not satisfy the requirement.
   • CCloudFront access logs on the endpoint is incorrect because this option does not satisfy the requirement.
4. Question 4Monitoring and Security

   A dashboard must show p90 and p99 ModelLatency for a SageMaker endpoint over time. Which CloudWatch capability provides percentile statistics natively?

   • ACloudWatch percentile statistics (e.g., p90, p99) on the metricCorrect
   • BOnly Average and Sum are available in CloudWatch
   • CPercentiles are only available in X-Ray and nowhere else
   • DPercentiles require exporting raw logs to Redshift first
   ✓ Correct answer: A

   CloudWatch integrates with SageMaker to collect endpoint metrics and logs. This integration enables real-time dashboards, alarms, and automated alerting without manual log aggregation. Logs capture training output for centralized troubleshooting.

   Why the other options are wrong

   • BOnly Average and Sum are available in CloudWatch is incorrect because this option does not satisfy the requirement.
   • CPercentiles are only available in X-Ray and nowhere else is incorrect because X-Ray does not capture API events.
   • DPercentiles require exporting raw logs to Redshift first is incorrect because this option does not satisfy the requirement.
5. Question 5Monitoring and Security

   An organization wants to govern and audit which foundation models, data sources, and ML configurations are used across many accounts, enforcing guardrails centrally. Which AWS capability supports central multi-account governance and policy enforcement?

   • AAmazon SQS standard queues
   • BAmazon CloudFront distributions
   • CA single shared IAM user across all accounts
   • DAWS Organizations with service control policies (SCPs)Correct
   ✓ Correct answer: D

   This correctly addresses the requirement following AWS best practices.

   Why the other options are wrong

   • AAmazon SQS standard queues is incorrect because SQS is a queuing service; not applicable.
   • BAmazon CloudFront distributions is incorrect because this option does not satisfy the requirement.
   • CA single shared IAM user across all accounts is incorrect because this option does not satisfy the requirement.
6. Question 6Deployment and OrchestrationSelect all that apply

   An organization wants to reduce inference cost for a mix of workloads: one endpoint has steady traffic, another is sporadic, and a third processes huge offline batches nightly. Which TWO mappings are the most cost-appropriate? (Choose TWO)

   • AUse Serverless Inference for the sporadic, low-and-intermittent-traffic workloadCorrect
   • BUse Batch Transform for the steady low-latency real-time workload
   • CUse a always-on GPU real-time endpoint for the nightly offline batch job
   • DUse Batch Transform for the nightly large offline scoring workloadCorrect
   ✓ Correct answer: A, D

   Use Batch Transform for the nightly large offline scoring workload Right-sizing instances, auto scaling, serverless patterns, and batching are fundamental cost-optimization strategies for ML inference workloads. They allow matching compute to actual demand while maintaining latency, avoiding overprovisioning. These practices form the foundation of responsible ML operations.

   Why the other options are wrong

   • BUse Batch Transform for the steady low-latency real-time workload is incorrect because this option does not satisfy the requirement.
   • CUse a always-on GPU real-time endpoint for the nightly offline batch job is incorrect because this option does not satisfy the requirement.
7. Question 7Monitoring and SecuritySelect all that apply

   A FinOps and ML platform team wants ongoing, data-driven recommendations to cut AWS cost across their ML stack. Which TWO AWS tools provide actionable cost-optimization insights? (Choose TWO)

   • AAWS Shield Advanced for DDoS protection
   • BAmazon Macie for sensitive-data discovery
   • CAWS Cost Explorer with Savings Plans/Reserved purchase recommendationsCorrect
   • DAWS Compute Optimizer for right-sizing recommendations on supported compute resourcesCorrect
   ✓ Correct answer: C, D

   AWS Compute Optimizer for right-sizing recommendations on supported compute resources This correctly addresses the requirement following AWS best practices.

   Why the other options are wrong

   • AAWS Shield Advanced for DDoS protection is incorrect because this option does not satisfy the requirement.
   • BAmazon Macie for sensitive-data discovery is incorrect because this option does not satisfy the requirement.
8. Question 8Deployment and OrchestrationSelect all that apply

   A SageMaker Pipeline that retrains and deploys a model fails intermittently at the processing step with transient S3 throttling and occasional spot-capacity interruptions. Which TWO changes improve reliability? (Choose TWO)

   • Aadd retry policies with backoff on the pipeline/processing stepCorrect
   • BEnable checkpointing or fall back to on-demand instances for steps sensitive to interruptionCorrect
   • CGrant the pipeline role administrator access
   • Dremove the evaluation/condition step to speed things up
   ✓ Correct answer: A, B

   Enable checkpointing or fall back to on-demand instances for steps sensitive to interruption This correctly addresses the requirement following AWS best practices.

   Why the other options are wrong

   • CGrant the pipeline role administrator access is incorrect because this option does not satisfy the requirement.
   • Dremove the evaluation/condition step to speed things up is incorrect because this option does not satisfy the requirement.
9. Question 9Deployment and Orchestration

   An inference endpoint that worked in testing fails in production with image pull errors, and logs show it cannot reach Amazon ECR because the endpoint runs in a locked-down VPC. What is the correct fix?

   • ASwitch the model artifact format to ONNX
   • BIncrease the endpoint's autoscaling maximum
   • Cadd ECR (and ECR API/DKR) interface VPC endpoints, plus an S3 gateway endpoint for the image layersCorrect
   • DMake the ECR repository public to everyone on the internet
   ✓ Correct answer: C

   Running in private VPC restricts data through security groups. Private subnets and VPC endpoints keep communication in AWS backbone. This architecture meets residency and compliance requirements.

   Why the other options are wrong

   • ASwitch the model artifact format to ONNX is incorrect because this option does not satisfy the requirement.
   • BIncrease the endpoint's autoscaling maximum is incorrect because this option does not satisfy the requirement.
   • DMake the ECR repository public to everyone on the internet is incorrect because this option does not satisfy the requirement.
10. Question 10Monitoring and Security

    A team must demonstrate continuous compliance that every SageMaker notebook instance has direct internet access disabled, and be alerted with automatic remediation when one drifts. Which service evaluates and remediates this configuration rule?

    • AAmazon SQS
    • BAWS Global Accelerator
    • CAmazon CloudFront
    • DAWS Config with a managed/custom rule and an automatic remediation actionCorrect
    ✓ Correct answer: D

    This correctly addresses the requirement following AWS best practices.

    Why the other options are wrong

    • AAmazon SQS is incorrect because SQS is a queuing service; not applicable.
    • BAWS Global Accelerator is incorrect because this option does not satisfy the requirement.
    • CAmazon CloudFront is incorrect because this option does not satisfy the requirement.

AWS MLA-C01 practice exam FAQ