AWS Machine Learning Specialty (MLS-C01)

Create data repositories, ingest and transform data, and ensure data quality for ML workloads on AWS.

• Amazon S3: data lake patterns, partitioning, storage classes for ML datasets
• AWS Glue: ETL jobs, crawlers, Data Catalog, Glue DataBrew for data prep
• Amazon Kinesis: streaming data ingestion for real-time ML
• AWS Lake Formation: governed data lakes, permissions, cross-account sharing
• Amazon Redshift: feature extraction from data warehouse for ML
• Data formats for ML: Parquet, ORC, CSV, RecordIO, TFRecord, libSVM
• Feature engineering pipelines: AWS Glue, EMR Spark for large-scale transforms
• Data labeling: Amazon SageMaker Ground Truth, labeling workforce options
• Data validation and quality checks: AWS Deequ, SageMaker Clarify data checks
• Data versioning and lineage with SageMaker ML Lineage Tracking

Free Data Engineering practice questions →Data Engineering study guide →

Sanitize, prepare, and analyze data using AWS services to understand distributions, correlations, and feature relationships.

• Amazon SageMaker Studio: notebooks for EDA, built-in visualizations
• Amazon SageMaker Data Wrangler: no-code data prep, feature transforms, quality reports
• Handling missing data: imputation strategies in SageMaker Processing
• Feature scaling: normalization, standardization for SageMaker algorithms
• Encoding categorical variables: one-hot, ordinal encoding in SageMaker
• Dimensionality reduction: PCA built-in algorithm in SageMaker
• Detecting and handling outliers and class imbalance (SMOTE, class weights)
• Correlation analysis and feature selection techniques
• AWS Glue DataBrew: 250+ data transforms, profiling, quality rules
• Amazon Athena: SQL-based EDA on S3 data at scale
• Amazon QuickSight: BI visualizations for data exploration

Free Exploratory Data Analysis practice questions →Exploratory Data Analysis study guide →

Select, train, evaluate, and tune ML models using Amazon SageMaker built-in algorithms and custom frameworks.

• SageMaker built-in algorithms: XGBoost, Linear Learner, K-Means, Random Cut Forest, BlazingText, DeepAR
• SageMaker built-in image algorithms: Image Classification, Object Detection, Semantic Segmentation
• SageMaker built-in NLP: Sequence-to-Sequence, LDA, Neural Topic Model
• Custom training containers: bring your own script with TensorFlow, PyTorch, MXNet
• SageMaker Training: training job configuration, instance types, distributed training
• Hyperparameter tuning: SageMaker Automatic Model Tuning (Bayesian optimization)
• Model evaluation metrics: RMSE, AUC-ROC, precision-recall, confusion matrix
• Overfitting vs underfitting: regularization (L1/L2), dropout, early stopping
• Cross-validation and train/validation/test splits in SageMaker
• SageMaker Experiments: tracking training runs and comparing models
• SageMaker Clarify: bias detection, feature importance (SHAP), explainability
• Choosing the right algorithm: classification, regression, clustering, anomaly detection
• Transfer learning with SageMaker JumpStart pre-trained models
• Amazon Bedrock: foundation models for generative AI use cases
• Reinforcement learning with SageMaker RL and Ray

Free Modeling practice questions →Modeling study guide →

Deploy, monitor, retrain, and secure ML models in production using SageMaker and AWS MLOps services.

• SageMaker model deployment: real-time endpoints, serverless inference, async inference, batch transform
• SageMaker multi-model endpoints and multi-container endpoints
• SageMaker Pipelines: MLOps CI/CD for ML workflows
• SageMaker Model Registry: versioning, approval workflows, deployment stages
• SageMaker Model Monitor: data drift, model quality, bias drift, feature attribution drift
• SageMaker A/B testing: production variants, traffic splitting
• Auto Scaling for SageMaker endpoints: target tracking, step scaling
• SageMaker Feature Store: online and offline feature storage, time-travel
• Cost optimization: Spot training instances, right-sizing endpoints, Savings Plans
• Security: VPC isolation for SageMaker, IAM roles, KMS encryption, PrivateLink
• SageMaker Edge Manager: deploying models to edge devices
• Model retraining triggers: concept drift, scheduled retraining pipelines
• AWS Step Functions for ML orchestration
• Amazon EventBridge and Lambda for event-driven ML pipelines

Free Machine Learning Implementation and Operations practice questions →Machine Learning Implementation and Operations study guide →