AI Wisdom

Databricks associate level

Databricks Certified Generative AI Engineer Associate

Code
GenAI Engineer
Length
90 minutes
Price
$200
Valid for
2 years, renewed by taking the current exam
Study time, our estimate
40 to 60 hours

A focused, practical RAG and agents exam that is the natural pick for data engineers moving into AI. It assumes your data already lives in the lakehouse and asks you to build on top of it, with governance built in rather than bolted on.

Take it if

Your company runs on Databricks and you want to build AI on that data.

Skip it if

You do not use Databricks. The platform specifics will not carry over.

Before you start

Check yourself against this list first. If something is missing, close that gap before you pay for an exam.

  • Six months of hands-on GenAI workDatabricks recommends this.
  • A Databricks workspaceYou need one for practice.
  • Python and SQLExpected for working on the platform.

Helpful first: DEA-C01

All exam details
Format
45 scored multiple-choice and multiple-selection questions, plus up to 10 unscored
Length
90 minutes
Price
$200
Valid for
2 years, renewed by taking the current exam
Delivery
Online or test center, proctored
Aids
None allowed
Experience
6+ months of hands-on generative AI work
If you fail
Wait 14 days

What the exam covers

  • Design Applications14%
  • Data Preparation14%
  • Application Development30%
  • Assembling and Deploying Apps22%
  • Governance8%
  • Evaluation and Monitoring12%

What to focus on

  • Application Development is the largest section: chains, tools, prompts and agents.
  • Assembling and deploying with Model Serving and MLflow.
  • Chunking and loading documents into AI Search, which Databricks called Vector Search until recently.
  • Unity Catalog for governing data, models and tools.
  • Evaluating quality with MLflow and monitoring in production.
Everything on one mapEvery topic and key idea in this certification.
mindmap
  root((GenAI Engineer))
    Design and data
      RAG first
      Chunking
      Synced index
    Application development
      Agents
      Tools
      Structured output
    Assemble and deploy
      MLflow
      Model Serving
      External models
    Governance and quality
      Unity Catalog
      Tracing
      Evaluation
A lakehouse RAG pathCheck product names against the exam guide. Databricks renamed Vector Search to AI Search.
flowchart LR
  A["Documents"] --> B["Chunk into Delta tables"]
  B --> C["AI Search index"]
  C --> D["Agent or chain"]
  D --> E["Log with MLflow"]
  E --> F["Model Serving endpoint"]
  F --> G["Monitor quality"]
  G --> D

Lesson 1 of 4, 14% + 14% of the exam

Design and data

A little over a quarter of the exam.

RAG first
Use retrieval when answers must come from your documents.
Chunking
Keep related text together.
Synced index
An AI Search index is built from a source table.
Building an index from a tableThe index follows the source table.
flowchart LR
  A["PDFs"] --> B["Parse and chunk"]
  B --> C["Delta table"]
  C --> D["AI Search index"]
  E["Question"] --> D
  D --> F["Relevant chunks"]

Common trap

Fixed character chunking on structured documents.

Try it yourself

Create an index from a Delta table.

Official docs: Databricks AI Search

Lesson 2 of 4, 30% of the exam

Application development

The largest section at 30%.

Agents
Databricks documents Agent Bricks as its agent developer platform.
Tools
Live data needs a tool, not an index.
Structured output
Use a schema and validate it.
Index or tool?Documents and live data take different routes.
flowchart TD
  A{"What does the agent need?"} -->|"Knowledge from documents"| B["AI Search index"]
  A -->|"Live data or an action"| C["Tool call"]
  B --> D["Agent"]
  C --> D

Common trap

Expecting a vector index to hold live order status.

Try it yourself

Build an agent with one tool.

Official docs: Develop agents on Databricks

Lesson 3 of 4, 22% of the exam

Assemble and deploy

Over a fifth of the exam.

MLflow
Logs and traces your app.
Model Serving
Exposes models as a REST API.
External models
Can sit behind governed endpoints.
From code to endpointEach step adds traceability.
flowchart LR
  A["Agent code"] --> B["Log with MLflow"]
  B --> C["Register in Unity Catalog"]
  C --> D["Model Serving"]
  D --> E["REST API"]

Common trap

Deploying from a notebook with no versioning.

Try it yourself

Log an app and deploy it to an endpoint.

Official docs: Model ServingAgent observability and quality

Lesson 4 of 4, 8% + 12% of the exam

Governance and quality

A fifth of the exam.

Unity Catalog
The governance layer for data and AI.
Tracing
A complete record of each agent run.
Evaluation
Measure from the first trace through production.
Everything governed in one placeUnity Catalog treats each asset as a securable object.
flowchart TD
  A["Unity Catalog"] --> B["Tables"]
  A --> C["Models"]
  A --> D["Functions"]
  A --> E["Indexes"]

Common trap

Enforcing access in prompts.

Try it yourself

Restrict a model to one group and trace a request.

Official docs: What is Unity Catalog?Agent observability and quality

Study plan

  1. Week 1: data

    Parse and chunk documents into Delta tables, and sync them to an AI Search index.

  2. Week 2: build

    Build a RAG app and a tool-calling agent, and log them with MLflow.

  3. Week 3: deploy and govern

    Register the model in Unity Catalog, deploy to Model Serving, and set up access.

  4. Week 4: evaluate

    Run an evaluation and add monitoring. Then take the mock below.

Exam day and after

  • Databricks says related training is highly recommended, though there is no formal prerequisite.
  • Exams can include unscored items, which do not affect your result.
  • Databricks does not state a pass mark on the certification page.
Test yourself8 practice questions, scored by exam topic.
Start the mock exam

Keep going

Exam facts checked on October 31, 2026 against Databricks certification page. Lessons are our own explanations, so use the official docs as the final word, and confirm price and dates before booking.