The Databricks Certified Generative AI Engineer Associate exam validates that you can design, build, deploy, and evaluate generative AI applications on Databricks. It is a 45-question, 90-minute, online-proctored exam that costs US$200, spanning six domains and 56 objectives, with no published passing score.
Last updated August 2026.
The exam is 45 scored questions in 90 minutes, delivered online with a proctor. Questions are multiple-choice or multiple-selection, and some ask you to pick two answers. Registration is US$200, there is no formal prerequisite, and the certification is valid for two years, after which you retake the full exam to recertify.
The guide organizes the material into six domains spanning 56 objectives. Because 56 objectives map onto 45 scored questions, the exam samples objectives rather than testing every one, so broad coverage beats betting on a few favourites.
| Domain | Objectives | Share of objectives |
|---|---|---|
| Design Applications | 6 | ~11% |
| Data Preparation | 8 | ~14% |
| Application Development | 13 | ~23% |
| Assembling and Deploying Applications | 15 | ~27% |
| Governance | 4 | ~7% |
| Evaluation and Monitoring | 10 | ~18% |
Objective counts are transcribed from the official Databricks Generative AI Engineer Associate exam guide (the edition covering the live exam as of 18 March 2026). Databricks does not publish per-domain percentages or a passing score; the share column is derived from objective counts and shown only to indicate relative emphasis.
This is an applied exam: it rewards design and trade-off judgment across the Databricks generative AI stack, not term recognition. The two largest domains, Application Development and Assembling and Deploying Applications, center on building and shipping real LLM applications.
Chunking, embedding, and retrieval quality with Mosaic AI Vector Search, backed by chunk tables and indexes governed in Unity Catalog. The exam expects you to measure retrieval quality before tuning parameters like chunk size, overlap, hybrid search, or re-ranking, not to change them on intuition.
Authoring chains and agents with the Mosaic AI Agent Framework and MLflow, plus the prebuilt Agent Bricks options (Knowledge Assistant, Multiagent Supervisor, Information Extraction). Prompts are versioned and promoted with MLflow prompt versions and aliases, never by merging a git branch or overwriting a table.
Serving models through Foundation Model APIs and Mosaic AI Model Serving, fronted by Mosaic AI Gateway for rate limiting and usage and inference tracking. Endpoints are called from an application backend using the app's own credentials: putting tokens or API keys in browser JavaScript is always the wrong answer.
Governing data, indexes, and models through Models in Unity Catalog, exposing conversational analytics with Genie Spaces, and measuring quality with MLflow (mlflow.genai.evaluate, scorers, and LLM-as-a-judge). Strong answers pair rubrics with calibrated human judgment rather than treating a single LLM judge as ground truth.
Databricks lists no required prerequisite, but recommends roughly six months of hands-on experience and its Academy "Generative AI Engineering with Databricks" learning path. Treat that as the floor: the exam assumes you have actually built a retrieval or agent application, not just read about one.
A practical approach:
Because Databricks does not publish a cut score, the honest target is scoring consistently well on realistic mock exams rather than chasing a specific percentage.
The exam has 45 scored questions with a 90-minute time limit. Some unscored questions can also appear without being identified, with extra time factored in, so the number presented may be higher than 45.
Databricks does not publish an official passing score for this exam. Because no cut score is public, aim to score consistently well on full-length practice exams rather than targeting a specific percentage.
No prerequisite is required, but Databricks recommends about six months of hands-on generative AI experience and its Academy "Generative AI Engineering with Databricks" path. It is associate level but still assumes you have built real applications.
Registration is US$200, and the exam is delivered online with a proctor. The certification is valid for two years, after which you take the full exam again to recertify.
Most candidates need a few weeks of focused study on top of some hands-on Databricks experience. Plan for more time if Mosaic AI Vector Search, the Agent Framework, and MLflow evaluation are new to you, and practice on timed exams.
Certify's Generative AI Engineer Associate course covers all 56 objectives across 55 mobile-first lessons, with 7 timed practice exams. The first unit is free to start.
Start the course →