AWS Machine Learning Engineer: prepare and operate models

Objectives and progression

Independent preparation for the MLA-C02 syllabus, currently in English beta, with eight modules mapped to the four official domains. Learn to prepare data, avoid leakage, control training and experiments, evaluate RAG, choose inference, promote versions, and operate models and agents securely. Fictional cases cover lost valid records, contaminated evaluation, Spot interruptions, unsupported retrieval, shadow versus A/B confusion, timeouts after external effects, delayed labels, and dependencies in isolated containers. Includes primary sources, option explanations, progress, and an internal 32-decision assessment in 60 minutes. It is not a full mock or a substitute for applied experience.

Audience: ML/MLOps, development, data, APS, and project professionals preparing and operating ML and AI solutions.

Prerequisites: AWS, IAM, data, ML, and pipeline fundamentals; programming familiarity helps. This path requires no paid resources.

380 estimated study minutes

• Prepare usable data and detect failures before feeding training or retrieval.
• Maintain training-serving consistency and evaluate genuinely independent data.
• Control criteria, versions, and resources before increasing training complexity.
• Choose customization and measure the retrieval-generation chain.
• Choose the serving mode and limit exposure when promoting a model.
• Automate promotion and recovery without losing evidence of what ran.
• Distinguish infrastructure health, data change, and outcome quality.
• Apply appropriate permissions and isolation and prepare auditable operations.

Modules

1. Ingestion, storage, and quality
2. Features, splits, and leakage
3. Training, tuning, and reproducible experiments
4. Foundation models and RAG quality
5. Inference and controlled change
6. Pipelines, versions, and agent state
7. Monitoring, drift, and useful cost
8. Security and operational readiness

Continue learning

Data / AI

• AWS Certified AI Practitioner
• AWS Certified Solutions Architect Associate
• Python
• SQL
• Application Production Support

References and version

MLA-C02

• AWS Certified Machine Learning Engineer Associate transition and beta overview · 2026-10-01
• MLA-C02 exam guide · 2026-10-01
• MLA-C01 and MLA-C02 comparison · 2026-10-01
• MLA-C02 update announcement2026-07-14 · 2026-10-01
• MLA-C02 in-scope services · 2026-10-01
• Data Preparation for ML and AI · 2026-10-01
• ML Model and Foundation Model Development · 2026-10-01
• Deployment and Orchestration of ML and AI Workflows · 2026-10-01
• Operating Monitoring and Securing ML and AI Solutions · 2026-10-01
• SageMaker Feature Store · 2026-10-01
• SageMaker Pipelines · 2026-10-01
• SageMaker Model Registry · 2026-10-01
• SageMaker inference options · 2026-10-01
• SageMaker shadow variants · 2026-10-01
• SageMaker deployment guardrails · 2026-10-01
• SageMaker Model Monitor · 2026-10-01
• Model Monitor availability change · 2026-10-01
• SageMaker model quality monitoring · 2026-10-01
• SageMaker automatic model tuning · 2026-10-01
• SageMaker managed Spot training · 2026-10-01
• Managed MLflow on SageMaker · 2026-10-01
• AWS Glue Data Quality · 2026-10-01
• Amazon Bedrock Knowledge Bases · 2026-10-01
• Amazon Bedrock Guardrails · 2026-10-01
• Amazon Bedrock evaluations · 2026-10-01
• Amazon Bedrock data protection · 2026-10-01
• Amazon Bedrock Prompt management · 2026-10-01
• Amazon Bedrock model customization · 2026-10-01
• Amazon Bedrock IAM · 2026-10-01
• Amazon Bedrock AgentCore overview · 2026-10-01
• SageMaker production and shadow variants · 2026-10-01
• SageMaker shadow-test limits · 2026-10-01
• AWS MLOps splits and data leakage · 2026-10-01
• SageMaker training VPC access · 2026-10-01
• SageMaker network isolation · 2026-10-01
• SageMaker execution roles · 2026-10-01