AWS Certified Machine Learning Engineer Associate (MLA‑C02) Practice Exams
About the AWS MLA-C02 exam
Exam at a glance
MLA-C02 is the updated version of AWS's associate-tier Machine Learning Engineer exam. It validates that you can build, operationalize, deploy, and maintain AI and ML solutions and pipelines on AWS, working with both traditional ML models and foundation models (FMs).
What's new in MLA-C02
Generative AI now runs through every domain. Alongside Amazon SageMaker AI, the exam guide adds vector databases and document chunking for Retrieval Augmented Generation (RAG), foundation model selection and customization in Amazon Bedrock, Bedrock Knowledge Bases, Guardrails and Prompt Management, and agent deployment, state and observability with Amazon Bedrock AgentCore. AWS also dropped several MLA-C01 skills, including SageMaker Neo edge optimization and model size reduction, so your study time goes where the exam now tests.
How it differs from AIF-C01 and AIP-C01
- AIF-C01 (AI Practitioner): foundational AI and ML concepts, terminology and AWS service overviews. An on-ramp for non-engineering roles.
- MLA-C02 (this exam): hands-on ML and generative AI engineering on AWS. Data preparation, model and FM development, deployment, MLOps, monitoring and security.
- AIP-C01 (Generative AI Developer Professional): professional-tier Gen AI application development with Amazon Bedrock, RAG, agents and prompt engineering.
Domain weighting
- Data Preparation for ML and AI: 28%
- ML Model and Foundation Model (FM) Development: 24%
- Deployment and Orchestration of ML and AI Workflows: 24%
- Operating, Monitoring, and Securing ML and AI Solutions: 24%
Who it's for
AWS targets candidates with at least 1 year of experience using Amazon SageMaker AI, Amazon Bedrock and other AWS services for ML engineering, plus at least 1 year in a related role such as backend software developer, DevOps developer, data engineer or data scientist. You should have worked with both traditional ML and generative AI.
Why take this certification
- The current ML engineering credential on AWS. Per AWS, GA delivery of MLA-C02 begins on January 14, 2027, and MLA-C01 is retired in all languages on that date.
- Traditional ML and generative AI in one exam. You prove you can ship both a SageMaker AI model and a Bedrock-based RAG or agent workload, the mix most ML teams now run.
- Production-ready skills. The exam tests building, deploying, monitoring and securing real workloads: feature stores, model registries, drift detection, guardrails and cost controls.
What you'll learn in the MLA-C02 exam
MLA-C02 validates that you can take ML and generative AI workloads from raw data to a monitored, secured production system on AWS. Questions are scenario-driven: a company or ML engineer, a stated setup, and requirements such as latency, cost, compliance or operational overhead that only one option meets.
Domain 1: Data Preparation for ML and AI (28%)
- Collect and store data: extract from Amazon S3, EBS, EFS, RDS, DynamoDB and OpenSearch Service; stream with Amazon Kinesis, Apache Flink and Apache Kafka; choose Parquet, JSON, CSV or ORC; configure vector databases (OpenSearch Service, RDS with pgvector, S3); ingest into SageMaker Feature Store.
- Transform and engineer features: AWS Glue, Glue DataBrew, Spark on Amazon EMR and SageMaker Data Wrangler; scaling, binning and log transforms; embedding models; tokenization; RAG chunking and metadata extraction; masking and anonymization; data for FM fine-tuning, continued pre-training and distillation.
- Validate data and manage bias: AWS Glue Data Quality and DataBrew checks, labeling, bias metrics across numeric, text and image data, class imbalance, prompt-response pair validation, content safety screening, outliers, imputation and deduplication.
Domain 2: ML Model and Foundation Model (FM) Development (24%)
- Choose modeling approaches: select FMs in Amazon Bedrock, pick fine-tuning strategies, weigh AWS AI services such as Amazon Textract, Rekognition, Comprehend and Transcribe against custom models, and select RAG architecture patterns on performance, latency and cost.
- Train and customize: SageMaker AI built-in algorithms, script mode, automatic model tuning, early stopping and distributed training, overfitting and catastrophic forgetting, ensembles, prompt engineering, fine-tuning, and tuning retrieval components and embedding models.
- Evaluate: reproducible experiments with MLflow on SageMaker AI, Amazon Bedrock evaluations and Prompt Management; baselines and drift detection; shadow variants; explainability; BLEU, ROUGE and BERTScore; human-in-the-loop review, LLM-as-a-judge and RAG retrieval accuracy.
Domain 3: Deployment and Orchestration of ML and AI Workflows (24%)
- Deployment infrastructure: compute targets, real-time and batch inference, multi-model and multi-container endpoints, FM deployment options, Amazon Bedrock Custom Model Import for models built outside AWS, agents and agent communication protocols, retrieval strategies and reranking.
- Provision resources: on-demand versus provisioned capacity, containers, SageMaker AI endpoints inside a VPC, auto scaling metrics, Amazon Bedrock knowledge bases, retrieval pipelines, agent state management, GPU scaling and agentic workflow infrastructure.
- MLOps and CI/CD: deployment strategies and rollback with AWS CodePipeline, CodeBuild, CodeDeploy, CodeCommit and CodeConnections; retraining triggers; versioning with SageMaker Model Registry and MLflow; Bedrock Prompt Management; agent deployment pipelines, prompt testing and knowledge base refresh cycles.
Domain 4: Operating, Monitoring, and Securing ML and AI Solutions (24%)
- Monitor: Amazon CloudWatch generative AI observability, Amazon Bedrock Model Evaluation, drift detection pipelines, A/B testing, and agent issues such as coordination failures, truncated streaming and tool failures.
- Optimize cost and performance: inference instance families, CloudWatch, Amazon Bedrock AgentCore Observability and AWS X-Ray, dashboards, AWS Cost Explorer and AWS Budgets, purchasing options, FM token usage, embedding costs and vector storage.
- Secure: Amazon Inspector and Amazon CodeGuru scans in CI/CD, least-privilege IAM, AWS CloudTrail and AWS Config, VPC isolation, Amazon Bedrock API keys versus IAM credentials, and Amazon Bedrock Guardrails for sensitive data and responsible AI.
How the practice exams help
Every question follows the scenario style AWS uses, with stated requirements that only the correct option satisfies. Explanations cover why the right answer works and why each distractor fails, with links to AWS documentation, so you learn the tradeoffs (a Bedrock knowledge base versus a custom retrieval pipeline, asynchronous versus batch inference) rather than memorizing answers.
How to prepare for the MLA-C02 exam
A strong MLA-C02 plan combines the official exam guide, structured study, hands-on practice on SageMaker AI and Amazon Bedrock, and exam-style scenario drills. Recommended approach:
- Start with the free sample. Take the 10 free questions above to benchmark where you stand. Read every explanation, including why each wrong answer is wrong.
- Study with the online study guide. Work through the MLA-C02 study guide, organized by the four exam domains. It explains each topic, the AWS services involved, and the decisions the exam asks you to make.
- Build hands-on. Train and tune a model with SageMaker AI, register it in Model Registry and deploy it to an endpoint with Model Monitor attached. Then create an Amazon Bedrock knowledge base over documents in Amazon S3, add a guardrail, and test prompts with Prompt Management.
- Drill with the premium practice exams. The complete course gives you 24 premium practice exams with 1,560 questions across all four domains, each with detailed explanations and AWS documentation references. Use your results to target the domains that need the most work.
If you studied for MLA-C01
Most of your SageMaker AI knowledge carries over. Put extra time into the new generative AI tasks: vector databases and chunking for RAG, Bedrock model selection and evaluations, Knowledge Bases, Guardrails, Prompt Management, and agents with Amazon Bedrock AgentCore. Skip the skills AWS removed, such as SageMaker Neo and model size reduction.
Official resources
Start with the official AWS Machine Learning Engineer Associate exam page and the MLA-C02 exam guide. The MLA-C01 and MLA-C02 comparison lists what changed. For depth, the SageMaker AI developer guide and the Amazon Bedrock user guide are the canonical references.