Everything for the Databricks Certified Generative AI Engineer Associate exam — the six domains, format, passing score, the Databricks toolchain it tests, and 240+ original practice questions with explanations.
Explore the course →This is Databricks' credential for engineers who build and ship LLM-enabled applications — most commonly a retrieval-augmented generation (RAG) pipeline or an LLM chain that runs on Databricks. Where many AI certifications stop at concepts, this one is explicitly engineering-focused.
The weights are your study budget — Application Development and Deployment together are 52% of the exam.
Translate business requirements into model tasks, decompose problems, and choose prompts, models, and chain components.
Chunking, embeddings, and building Vector Search indexes (Delta Sync vs Direct Access, managed vs self-managed embeddings).
Build the RAG chain, improve retrieval (reranking, hybrid search), and add agents, tools, guardrails, and prompt versioning.
Package with MLflow, register in Unity Catalog, and serve with the right endpoint mode (pay-per-token vs provisioned throughput).
Access control, PII handling, licensing, and safety with Unity Catalog and Mosaic AI Gateway.
Measure faithfulness and relevance with MLflow, then monitor live quality, cost, and drift.
Semantic retrieval over your embeddings; Delta Sync vs Direct Access indexes, managed vs self-managed embeddings.
Hosted models via pay-per-token or provisioned throughput, through a unified OpenAI-compatible API.
Deploy chains and agents as endpoints, with scale-to-zero, aliases, and traffic splitting.
Log and register models, Tracing, the Prompt Registry, and mlflow.genai.evaluate() for evaluation.
Governance: the catalog.schema.object namespace, permissions, and lineage.
Knowledge Assistant, Multi-Agent Supervisor, and Information Extraction; guardrails and rate limiting.
45 scored questions in 90 minutes (about two minutes each). Nearly every question is a short scenario with the deciding constraint inside it.
Databricks reports pass/fail with a domain-by-domain breakdown rather than a fixed percentage; roughly 70% is the commonly cited working target. Be solid across all six domains.
Design Applications (14%), Data Preparation (14%), Application Development (30%), Assembling & Deploying (22%), Governance (8%), and Evaluation & Monitoring (12%).
Vector Search, Mosaic AI Model Serving, Foundation Model APIs, MLflow, Unity Catalog, Mosaic AI Gateway, and Agent Bricks. Code is Python (some SQL).
Yes. It is engineering-focused: you should be comfortable reading Python and building a RAG pipeline or LLM chain on Databricks.
Yes. The exam guide, module lessons, and 240+ practice questions with explanations are free on the CtrlSkill learning platform.
Free exam guide, six module lessons, and 240+ practice questions with explanations.
See all resources →CtrlSkill is not affiliated with or endorsed by Databricks. Practice questions are original, created from the official exam-guide objectives, and are not actual exam questions. Fees, format, and objectives can change — always confirm on the official Databricks certification page before booking. Verified against Databricks documentation, September 2026.