The complete AWS Machine Learning Engineer Associate (MLA-C02) guide

What MLA-C02 is, and who it is for

MLA-C02 is an associate-level AWS certification for people who put machine learning into production, not just people who train models. The exam guide describes the target candidate as someone with at least one year of experience using Amazon SageMaker AI, Amazon Bedrock, and other AWS services for ML engineering, plus at least a year in a related role such as backend developer, DevOps developer, data engineer, or data scientist.

The defining feature of the C02 version is that it treats traditional ML and foundation models as one job. You are expected to move fluently between training an XGBoost model and standing up a Bedrock RAG application, because that is what the role now looks like.

What changed from MLA-C01

MLA-C01 was already an engineering-heavy ML credential. MLA-C02 keeps that spine and layers generative AI on top. The domain names were rewritten to say "ML and AI" instead of just "ML," and the weightings shifted slightly.

Weightings from the official MLA-C02 exam guide. New topics include Amazon Bedrock, embeddings and vector databases, RAG document preparation, agentic workflows, and guardrails.

The exam at a glance

MLA-C02 launched as a beta (exam code ME1-C02, English only, delivery from September 29, 2026). During the beta period, results are delayed and the standard pass or fail designation does not apply, per the exam guide. See the full cost, format and question types breakdown.

The four content domains

The exam is 28% Domain 1 and 24% each for Domains 2 through 4, so after the data domain there is no small area to skip. The full domains and weighting breakdown covers each in detail; in short:

• Domain 1: Data Preparation for ML and AI (28%). Ingesting and storing data, transformation and feature engineering, plus AI-specific preparation: embeddings, vector databases, and chunking documents for RAG.
• Domain 2: ML and Foundation Model Development (24%). Choosing a modeling approach across built-in algorithms and foundation models, training and fine-tuning, hyperparameter tuning, and evaluating both ML models and FM outputs.
• Domain 3: Deployment and Orchestration (24%). Selecting inference options (real-time, serverless, asynchronous, batch), autoscaling, infrastructure as code, and CI/CD with SageMaker Pipelines.
• Domain 4: Operating, Monitoring, and Securing (24%). Monitoring models and agents, cost and performance tuning, and securing ML and AI systems with IAM, network isolation, auditing, and guardrails.

How to prepare

The fastest blind-spot audit is free: read the exam guide's task statements straight through and mark the ones you could not explain to a colleague. Then close the gaps and drill scenario questions, because every exam item is a scenario with one best answer, not a definition recall.

• Read the task statements first. They are the literal source questions are written from, and they name the in-scope services per domain.
• Build one end-to-end solution. Train a model in SageMaker AI, deploy it to an endpoint, add monitoring, and script it. That single project touches all four domains.
• Do not skip the GenAI half. If you come from classic ML, Bedrock, embeddings, RAG, and guardrails are where the new marks live. If you come from app or data engineering, the modeling and evaluation topics are the runway.
• Practice at exam pace. 130 minutes for 65 questions is two minutes each, and multiple-response items eat time. Timed practice trains the pace as much as the content.

For a week-by-week schedule, see how long to prepare for MLA-C02.

Is MLA-C02 worth taking?

If your work is putting ML or GenAI into production on AWS, yes. MLA-C02 is one of the first associate-level exams to formally test foundation model and agentic workloads, so passing it signals current, operational skills rather than textbook ML theory. It pairs naturally with the AWS AI Practitioner (AIF-C01) if you want a foundational credential first, but MLA-C02 is the one that maps to the engineering job.

FAQ