AWS Certified Machine Learning Engineer - Associate Exam Prep

Covers the end-to-end preparation of data for ML workloads, including ingestion, storage, transformation, feature engineering, quality checks, bias handling, splitting, and labeling. This domain emphasizes selecting the right AWS services and data-processing patterns to produce reliable training and evaluation datasets.

• Task 1.1: Ingest and store data
• Data formats and ingestion mechanisms (CSV, JSON, Parquet, ORC, Avro, RecordIO)
• AWS storage options for ML workloads: Amazon S3 (Standard, Intelligent-Tiering, Glacier classes), Amazon EBS, Amazon EFS, Amazon FSx for Lustre (for high-throughput training reads)
• Task 1.2: Transform data and perform feature engineering
• Data cleaning, normalization, encoding (one-hot, target, ordinal), binning, imputation
• Feature engineering: aggregations, time-window features, embeddings, derived features
• Task 1.3: Ensure data integrity and prepare data for modeling

Key references: AWS MLA-C01 official exam guide · ExamPal shared topic tree

Covers selecting modeling approaches, training and refining models, and evaluating performance across common ML and NLP tasks. This domain emphasizes SageMaker training, tuning, transfer learning, and the use of appropriate metrics and analysis tools.

• Task 2.1: Choose a modeling approach
• ML problem framing: classification, regression, clustering, anomaly detection, recommendation, forecasting
• Algorithm selection: linear regression, logistic regression, XGBoost, k-means, RCF, DeepAR, BERT, neural networks
• Task 2.2: Train and refine models
• SageMaker training jobs (spot training, distributed training, pipe mode vs file mode)
• Hyperparameter tuning: SageMaker Automatic Model Tuning (Bayesian, random, grid, Hyperband)
• Task 2.3: Analyze model performance

Key references: AWS MLA-C01 official exam guide · ExamPal shared topic tree

Covers deployment choices, infrastructure scripting, workflow orchestration, and CI/CD for ML solutions. This domain includes SageMaker endpoint patterns, IaC tools, pipeline orchestration, and release strategies for safe model deployment.

• Task 3.1: Select deployment infrastructure based on existing architecture and requirements
• SageMaker endpoint types: real-time, serverless, asynchronous, batch transform
• Multi-model endpoints, multi-container endpoints, inference pipelines
• Task 3.2: Create and script infrastructure based on existing architecture and requirements
• Infrastructure as Code for ML: AWS CloudFormation, AWS CDK, SageMaker Projects
• SageMaker Pipelines for ML workflows (preprocessing → training → evaluation → deployment)
• Task 3.3: Use automated orchestration tools to set up continuous integration and continuous delivery (CI/CD) pipelines

Key references: AWS MLA-C01 official exam guide · ExamPal shared topic tree

Covers monitoring model and infrastructure health, optimizing cost and resource usage, and securing ML systems on AWS. This domain includes drift detection, endpoint observability, IAM, encryption, network isolation, documentation, secrets management, and compliance logging.

• Task 4.1: Monitor model performance and data quality
• SageMaker Model Monitor: data quality drift, model quality drift, bias drift, feature attribution drift
• Concept drift vs data drift detection patterns
• Task 4.2: Monitor and optimize infrastructure and costs
• Cost optimization: spot training, savings plans, right-sizing instance types
• AWS Cost Explorer, AWS Budgets for ML cost tracking
• Task 4.3: Secure AWS resources

Key references: AWS MLA-C01 official exam guide · ExamPal shared topic tree