AWS Certified Machine Learning Engineer Associate Training

This course prepares you for the MLA-C01 certification exam, covering all official exam domains and their approximate weightings:

Domain 1 Domain 1: Data Preparation for Machine Learning (ML) (28%)

• Ingest and store data from AWS sources (S3, EFS, FSx) and streaming sources (Kinesis, Apache Flink/Kafka); choose data formats (Parquet, JSON, CSV, ORC, Avro, RecordIO)
• Transform data and perform feature engineering (cleaning, encoding, scaling/normalization, binning) using SageMaker Data Wrangler, AWS Glue/Glue DataBrew, Spark on EMR
• Create and manage features using SageMaker Feature Store; validate and label data using SageMaker Ground Truth / Mechanical Turk
• Ensure data integrity: identify and mitigate bias (class imbalance, DPL) using SageMaker Clarify
• Apply data classification, anonymization, masking and encryption for compliance (PII, PHI, data residency)
• Prepare data for modeling (splitting, shuffling, augmentation) and configure data loading into training resources (EFS, FSx)

Domain 2 Domain 2: ML Model Development (26%)

• Choose a modeling approach: assess feasibility, select ML algorithms/AI services (Bedrock, Rekognition, Translate, Transcribe), consider interpretability and cost
• Use SageMaker built-in algorithms, script mode (TensorFlow/PyTorch), and JumpStart/Bedrock foundation models for fine-tuning
• Train and refine models: hyperparameter tuning (SageMaker AMT), regularization (dropout, L1/L2), prevent overfitting/underfitting/catastrophic forgetting
• Combine models via ensembling, stacking, boosting; reduce model size via pruning/compression/quantized data types
• Manage model versions using SageMaker Model Registry for repeatability and audits
• Analyze model performance: select/interpret evaluation metrics (F1, precision/recall, RMSE, ROC/AUC), create baselines, detect bias and convergence issues via SageMaker Clarify/Model Debugger

Domain 3 Domain 3: Deployment and Orchestration of ML Workflows (22%)

• Select deployment infrastructure: real-time/serverless/asynchronous endpoints vs batch inference, compute provisioning (CPU/GPU), containers, edge optimization (SageMaker Neo)
• Choose deployment orchestrator and target (SageMaker Pipelines, Airflow, SageMaker endpoints, ECS/EKS, Lambda) and deployment strategy (real time vs batch, blue/green, canary, linear)
• Create and script infrastructure as code (CloudFormation, AWS CDK) including containerization (ECR, EKS, ECS, bring-your-own-container) and SageMaker endpoint auto scaling
• Configure SageMaker endpoints within a VPC and deploy/host models using the SageMaker SDK
• Set up CI/CD pipelines with AWS CodePipeline, CodeBuild, and CodeDeploy, integrated with Git-based version control
• Automate orchestration of training/inference jobs (EventBridge rules, SageMaker Pipelines) and build automated tests plus retraining mechanisms

Domain 4 Domain 4: ML Solution Monitoring, Maintenance, and Security (24%)

• Monitor model inference: detect data/model drift and anomalies using SageMaker Model Monitor and SageMaker Clarify; monitor performance via A/B testing
• Monitor and optimize infrastructure and costs using CloudWatch, X-Ray, CloudTrail, Cost Explorer, Trusted Advisor, and resource tagging strategies
• Rightsize instances and troubleshoot latency/scaling/capacity issues using SageMaker Inference Recommender and AWS Compute Optimizer
• Optimize infrastructure costs via purchasing options (Spot, On-Demand, Reserved Instances, SageMaker Savings Plans)
• Secure AWS resources: configure least-privilege IAM roles/policies for ML systems and applications, including SageMaker Role Manager
• Build VPCs, subnets, and security groups to isolate ML systems; monitor, audit, and log ML systems for continued security and compliance