Databricks Certified Generative AI Engineer Associate

This part turns a business requirement into the structure of a GenAI application.

• Translating goals into the inputs and outputs of an AI pipeline
• Selecting model tasks and chain components
• Defining and ordering tools for multi-stage reasoning

Here you prepare the documents a RAG application will later answer from.

• Extracting content from source documents and filtering out noise
• A chunking strategy that fits document and model
• Writing chunks to Delta tables in Unity Catalog
• Indexes for semantic search

The largest exam domain: prompt, retrieval and model become a working application.

• Augmenting prompts with context from the user input
• Retrieval and LLM chains
• Choosing the LLM and embedding model to fit requirements
• Agents with MLflow and the Agent Framework

This part puts the application into service through Unity Catalog and Model Serving.

• A chain as a pyfunc model with pre- and post-processing
• Registering it in Unity Catalog with MLflow
• Creating and querying a Vector Search index
• Model Serving and what resources endpoints can access

Here you protect data and application against misuse and data leakage.

• Access rights with Unity Catalog
• Masking sensitive data as a guardrail
• Guardrails against malicious user input

Finally you measure how well the application works and what it costs.

• Evaluating agents with MLflow scoring and tracing
• Choosing metrics for a production LLM scenario
• Inference tables for usage and quality in operation
• Controlling LLM costs with Databricks features