The AWS Machine Learning Engineer Associate (MLA-C02) exam guide names a specific set of Amazon SageMaker AI features: Feature Store, Data Wrangler, built-in algorithms and script mode, automatic model tuning, MLflow on SageMaker AI, Model Registry and endpoints. Four tools that older study material leans on, Clarify, Debugger, Model Monitor and Ground Truth, are now closed to new customers, and the MLA-C02 guide names none of them.
Last updated October 2026.
SageMaker AI is still the center of the traditional ML half of MLA-C02, but the C02 guide is more specific than C01 was. These are the features its skills name directly:
| Feature | Where the guide names it |
|---|---|
| Feature Store | Domain 1: ingesting data and creating and managing features |
| Data Wrangler | Domain 1: transforming data, alongside AWS Glue, DataBrew and Spark on EMR |
| Built-in algorithms, script mode | Domain 2: applying algorithms and common ML libraries |
| Automatic model tuning (AMT) | Domain 2: hyperparameter optimization |
| MLflow on SageMaker AI | Domains 2 and 3: reproducible experiments and model versions |
| Model Registry | Domain 3: versioning for repeatability and audits |
| Endpoints and the SageMaker AI SDK | Domain 3: hosting in a VPC and deploying programmatically |
From the domain pages of the official MLA-C02 exam guide, checked 23 October 2026.
Deployment adds the choice between real-time, batch, asynchronous and serverless inference, auto scaling on endpoint metrics, and when to host on SageMaker AI versus Amazon ECS, Amazon EKS or AWS Lambda. Domains 3 and 4 also bring in Amazon Bedrock throughout, so SageMaker AI is half the platform story rather than all of it.
The SageMaker AI developer guide carries the same note on four features: "no longer open to new customers".
Status from each feature's page in the Amazon SageMaker AI developer guide, checked 23 October 2026.
The MLA-C02 guide does not name any of them. Its skills describe the task instead: "Label and annotate data", "Explain model outputs", "Debug model convergence issues", "Create model performance baselines and implement drift detection". Its in-scope services list names Amazon SageMaker AI as a whole, and none of the four as separate entries. Because they are part of SageMaker AI, a question can still mention one, so know what each did.
How to study them: learn the technique, not the console. SHAP attributions, pre- and post-training bias metrics, vanishing gradients and early stopping, baselines and statistical drift tests, and labeling quality are all testable without a specific tool. Expect the skill to be tested whichever way the question is worded.
Skill 4.1.1 names three things for monitoring models in production: Amazon CloudWatch generative AI observability, Amazon Bedrock Model Evaluation and drift detection pipelines. A drift detection pipeline compares live inputs and predictions with a training baseline on a schedule, and raises an alarm or triggers retraining when the gap grows. Operational problems, such as a failing job or a latency spike, belong to CloudWatch alarms and anomaly detection, not to drift detection.
If you studied with MLA-C01 material, the MLA-C01 vs MLA-C02 comparison lists what was added and removed, and the older SageMaker post for MLA-C01 shows how much the emphasis has moved.
The MLA-C02 guide names Feature Store, Data Wrangler, built-in algorithms and script mode, automatic model tuning, MLflow on SageMaker AI, Model Registry, and endpoints deployed with the SageMaker AI SDK. Inference options and endpoint auto scaling are also tested.
The guide does not name it. Clarify is closed to new customers, and the MLA-C02 guide tests the task, such as explaining model outputs and measuring bias. Because Clarify is part of SageMaker AI, which is in scope, know what it does, but learn SHAP attributions and bias metrics as techniques.
Not to new customers. The SageMaker AI developer guide states that Model Monitor is no longer open to new customers, and the MLA-C02 guide refers to drift detection pipelines instead.
The MLA-C02 guide does not name a labeling service; its skill is simply to label and annotate data. Ground Truth is closed to new customers, so study labeling quality and annotation workflows as concepts.
The Certified MLA-C02 course maps all four domains to 76 short chapters, with a practice exam per unit and three full-length mocks. Unit 1 is free.
Open the MLA-C02 course →