AWS Certified Machine Learning Engineer - Associate: Skills Measured

The AWS exam guide defines four MLA-C01 domains: Data Preparation for Machine Learning at 28%, ML Model Development at 26%, Deployment and Orchestration of ML Workflows at 22%, and ML Solution Monitoring, Maintenance, and Security at 24% of scored content. These official domains should replace local topic groupings when planning study.

This domain covers ingesting, storing, transforming, validating, and preparing data for modeling. Study S3 data storage, Glue and EMR processing, Athena queries, data quality checks, feature engineering, SageMaker Feature Store, and training-serving consistency. The exam often asks how to prepare reliable data for repeatable model work.

Model development covers selecting modeling approaches, training models, tuning hyperparameters, evaluating performance, and managing versions. Know SageMaker training jobs, tuning jobs, built-in algorithms where relevant, model registry concepts, evaluation metrics, Clarify for bias or explainability, and the tradeoffs between model quality, cost, and operational complexity.

This domain covers choosing deployment infrastructure, provisioning compute, configuring endpoints and auto scaling, and automating workflows. Compare batch transform, real-time endpoints, serverless inference, asynchronous inference, and multi-model endpoints. Also review SageMaker Pipelines, CI/CD integration, deployment guardrails, traffic shifting, and rollback-friendly delivery.

Monitoring and maintenance covers data quality, model quality, drift, infrastructure metrics, logs, retraining signals, and operational alerts. Security includes IAM roles, least privilege, encryption, VPC controls, data protection, and compliance-aware configuration. Cost optimization also matters because production inference and training choices affect ongoing spend.

AWS notes that full end-to-end ML architecture strategy, deep work across multiple ML domains, and quantization impact analysis are outside the expected target-candidate tasks. That helps frame study: focus on implementing and operating ML workflows on AWS rather than trying to master every advanced ML research topic.