Differentiate appropriate AWS storage and ingestion services for ML and AI workloads based on data requirements and compliance constraints.

Differentiate appropriate data formats based on access patterns and processing requirements.

Describe different methods for visualizing data.

Differentiate appropriate data transformation, feature engineering, and GenAI data preparation methods for a given dataset and modeling approach.

Describe document preparation for RAG applications including chunking strategies and metadata extraction.

Describe fundamental encoding techniques.

Differentiate data quality and bias risks across data modalities and appropriate mitigation strategies for ML and GenAI training data.

Explain data quality validation using AWS data quality tools.

Describe data labeling and annotation using AWS services.

Integrate multiple datasets by joining two cataloged tables in an AWS Glue Visual ETL job.

Orchestrate a crawler and an ETL job as a single AWS Glue Workflow using triggers.

Process data incrementally with AWS Glue job bookmarks.

Differentiate appropriate modeling approaches for a given business problem.

Differentiate FMs from Amazon Bedrock based on task requirements and cost.

Identify fine-tuning strategies for pre-trained FMs and when to use each.

Differentiate appropriate training approaches and configurations for ML models using SageMaker AI.

Define core elements in the model training process.

Explain SageMaker AI built-in algorithms and common ML libraries for training.

Differentiate appropriate techniques to refine and customize ML and GenAI models.

Explain techniques to prevent model overfitting, underfitting, and catastrophic forgetting

Describe model combination approaches, including ensemble methods such as boosting, bagging, and stacking.

Diagnose the performance of ML and GenAI models and systems.

Explain reproducible experiment tracking using MLflow on SageMaker AI and Amazon Bedrock evaluations.

Interpret model evaluation metrics such as model bias and model variance.

Differentiate appropriate deployment infrastructure for ML and AI workloads based on latency, cost, and scale requirements.

Differentiate appropriate compute environments and deployment targets for ML workloads.

Differentiate deployment orchestrators and multi-model or multi-container strategies.

Differentiate appropriate compute, storage, networking, and AI-specific resource configurations for ML and AI workloads.

Explain resource provisioning tradeoffs between on-demand and provisioned resources.

Describe compute provisioning automation.

Differentiate appropriate CI/CD pipeline components for automating ML and AI workloads.

Describe automated deployment strategies with rollback actions.

Explain AWS developer tools for CI/CD and their troubleshooting approaches.

Differentiate appropriate monitoring strategies for model performance, data quality, drift detection, agent behavior, and FM quality.

Describe how to monitor model performance and data quality in production using Amazon CloudWatch and Amazon Bedrock evaluations.

Explain how to detect anomalies or errors in data processing and model inference workflows.

Differentiate cost and performance tradeoffs for production ML and AI workloads.

Differentiate inference instance families to optimize performance and cost.

Describe troubleshooting and analysis tools for ML and AI workloads.

Diagnose security risks in an ML and AI system and the appropriate controls to mitigate them.

Explain how to secure CI/CD pipelines by checking for code and image vulnerabilities.

Describe least privilege access patterns for ML and AI artifacts.

Prepare data for ML and AI, including features, RAG documents, and fine-tuning datasets.

Choose, train, fine-tune, and evaluate ML models and foundation models.

Deploy and orchestrate ML and AI workflows using AWS services.