DATABRICKSDatabricks Certified Generative AI Engineer Associate
The Databricks Certified Generative AI Engineer Associate certification validates your ability to design, build, and deploy performant generative AI solutions on the Databricks platform. It is for engineers who develop LLM-enabled applications, including RAG systems and LLM chains, using Databricks tools like AI Search, Model Serving, MLflow, and Unity Catalog. Earning this credential demonstrates you can turn complex requirements into reliable, production-ready AI solutions.
644 practice questions · Updated 2026-07-30
6Domains
32Objectives
155Concepts
644Questions
GENERATIVE-AI-ENGINEER-ASSOCIATE Curriculum
Every domain, objective, and concept the GENERATIVE-AI-ENGINEER-ASSOCIATE exam measures.
- Prompt Formatting Instructions
- Structured Output Constraints
- Few-shot Formatting Examples
- System vs. User Prompt Roles
- Handling Ambiguity in Format Requests
- Testing and Iterating on Prompts
- Model Task Selection
- Business Requirement Analysis
- Task-Capability Mapping
- Constraint Consideration
- Chain components selection
- Input formatting
- Output parsing
- Chain orchestration
- Business Goal Analysis
- Input Specification
- Output Specification
- Alignment of Inputs and Outputs
- Multi-stage reasoning
- Tool gathering knowledge
- Tool taking actions
- Ordering tools in a pipeline
- Integration of tools in generative AI
- Agent Bricks overview
- Knowledge Assistant use cases
- Multiagent Supervisor use cases
- Information Extraction use cases
- Selection criteria for Agent Bricks
- Document structure analysis
- Chunking strategies
- Model context window constraints
- Chunk overlap and boundaries
- Evaluation of chunk quality
- Identify extraneous content
- Apply filtering techniques
- Preserve relevant information
- Evaluate filtering impact
- Document extraction packages
- Format-specific extraction
- Extraction workflow
- Chunking Text for Delta Lake
- Delta Lake Table Creation in Unity Catalog
- Writing Chunked Data to Delta Lake
- Sequence of Operations for Data Writing
- Unity Catalog Integration with Delta Lake
- Identify source document types
- Assess document quality
- Determine knowledge coverage
- Match documents to application needs
- Retrieval evaluation metrics
- Ground truth and relevance judgments
- Evaluating retrieval in RAG pipelines
- Tools for retrieval evaluation
- Chunking fundamentals
- Fixed-size chunking
- Recursive character text splitting
- Document-based chunking
- Semantic chunking
- Context-aware chunking
- Evaluating chunking strategies
- Purpose of re-ranking
- Re-ranking in the retrieval pipeline
- Initial retrieval vs. re-ranking
- Common re-ranking techniques
- Impact of re-ranking on performance
- Application attribute analysis for LLM selection
- Model card interpretation
- Model hub and marketplace navigation
- Task-specific model evaluation metrics
- Experiment result analysis for model selection
- Evaluation vs monitoring in Gen AI lifecycle
- Augmentation Fundamentals
- Key Field Extraction
- Term Extraction
- Intent Recognition
- Context Integration
- Baseline Response Understanding
- Prompt Adjustment Techniques
- Iterative Refinement
- Chunking strategy selection
- Embedding model context length selection
- Qualitative Response Assessment
- Safety Issue Identification
- Quality Issue Identification
- LLM Guardrails Overview
- Implementing Input Guardrails
- Implementing Output Guardrails
- Guardrail Evaluation and Iteration
- LangChain and similar tool selection
- MLflow for agentic systems
- Agent Framework for development
- Multi-agent system design
- Genie Spaces integration
- Conversational API for data retrieval
- RAG application components
- Mosaic AI Vector Search concepts
- Creating a Vector Search index
- Querying a Vector Search index
- Vector search configuration trade-offs
- Pyfunc chain with pre/post-processing
- Simple chain coding
- Serving LLM applications with Foundation Model APIs
- Batch inference with ai_query()
- Model Serving Endpoint Access Control
- MLflow Model Registration to Unity Catalog
- Persistent Datastore Configuration for Agent Memory
- CI/CD for Vector Search Index Updates
- Prompt Promotion Across Environments
- Component Testing for Agent Applications
- Prompt Version Control
- Prompt Lifecycle Management
- MCP server types
- MCP server integration
- Interface selection
- Interface development
- Purpose of masking in governance
- Dynamic data masking
- Static data masking
- Masking functions and patterns
- Role-based access control with masking
- Performance considerations of masking
- Identify malicious input types
- Apply input validation and sanitization
- Use output filtering and moderation
- Implement rate limiting and abuse detection
- Leverage guardrail frameworks and tools
- Design for secure prompt handling
- Identify data source licensing types
- Assess legal risks of data usage
- Apply compliance measures for data sourcing
- Identify problematic text in data sources
- Evaluate mitigation strategies
- Recommend a mitigation approach
- Quantitative evaluation metrics for LLMs
- Mapping metrics to LLM size and architecture
- Deployment scenario metric selection
- Inference Logging for RAG Applications
- Inference Tables for Live Endpoint Tracking
- Agent Monitoring for Deployed Agents
- AI Gateway Inference Tables
- AI Gateway Usage Tables
- Rate Limiting with AI Gateway
- MLflow scoring for agent evaluation
- MLflow tracing for agent evaluation
- Databricks custom Scorers for agents
- Custom Scorer configuration and integration
- SME feedback collection
- Incorporating SME feedback into agent improvement
- Databricks LLM cost control features
- Usage tracking and monitoring
- Cost optimization strategies
- Definition of ground truth in evaluation
- Evaluation judges requiring ground truth
- Judges not requiring ground truth
- Role of ground truth in judge selection
Ready to practice?Test your knowledge with exam-style questions or take an intelligent quiz tailored to your level.
Percentages reflect share of the current practice bank, not official exam weightings — no structured per-skill weight is published for GENERATIVE-AI-ENGINEER-ASSOCIATE, so none is invented.