Databricks Generative AI Engineer Associate
Design, build, govern, evaluate and deploy RAG apps and agents on Databricks. Lessons, exam-style practice, timed mocks and an AI tutor, mapped to every objective of the official exam guide.
₹399 12-month access · 7-day free trial
The exam
- Questions
- 45
- Time
- 90 minutes
- Passing score
- Not published by Databricks; this course targets ≥ 80% on unseen mocks
- Exam fee
- USD 200 (plus applicable taxes)
- Valid for
- 2 years (recertify by taking the current live exam)
- Delivery
- Online proctored
Official exam page · booked with Databricks
Every section of the exam guide
6 sections and 56 objectives, in the official weights. Each lesson is mapped to the objectives it covers.
Section 1 · Design Applications14%
What the exam asks
- Design a prompt that elicits a specifically formatted response
- Select model tasks to accomplish a given business requirement
- Select chain components for a desired model input and output
- Translate business use case goals into a description of the desired inputs and outputs for the AI pipeline
- Define and order tools that gather knowledge or take actions for multi-stage reasoning
- Determine how and when to use Agent Bricks (Knowledge Assistant, Multiagent Supervisor, Information Extraction) to solve problems
Lessons
- From Use Case to Prompt: Inputs, Outputs, Tasks and Formats
- Chains, Tools and Agent Bricks
Section 2 · Data Preparation14%
What the exam asks
- Apply a chunking strategy for a given document structure and model constraints
- Filter extraneous content in source documents that degrades quality of a RAG application
- Choose the appropriate Python package to extract document content from provided source data and format
- Define operations and sequence to write given chunked text into Delta Lake tables in Unity Catalog
- Identify needed source documents that provide necessary knowledge and quality for a given RAG application
- Use tools and metrics to evaluate retrieval performance
- Design retrieval systems using advanced chunking strategies
- Explain the role of re-ranking in the information retrieval process
Lessons
- Source Documents, Extraction and Filtering
- Chunking Strategies and Writing Chunks to Delta
- Advanced Retrieval: Chunking Designs, Evaluation and Re-ranking
Section 3 · Application Development30%
What the exam asks
- Select Langchain/similar tools for use in a Generative AI application
- Qualitatively assess responses to identify common issues such as quality and safety
- Select chunking strategy based on model & retrieval evaluation
- Augment a prompt with additional context from a user's input based on key fields, terms, and intents
- Create a prompt that adjusts an LLM's response from a baseline to a desired output
- Implement LLM guardrails to prevent negative outcomes
- Select the best LLM based on the attributes of the application to be developed
- Select an embedding model context length based on source documents, expected queries, and optimization strategy
- Select a model from a model hub or marketplace for a task based on model metadata/model cards
- Select the best model for a given task based on common metrics generated in experiments
- Utilize MLflow and Agent Framework for developing agentic systems
- Compare the evaluation and monitoring phases of the Gen AI application life cycle
- Enable multi-agent systems to leverage Genie Spaces or conversational API to retrieve data
Lessons
- Frameworks and tools for GenAI applications
- Augmenting and refining prompts
- Response quality, safety and guardrails across the lifecycle
- Choosing the right LLM
- Embedding context length and chunking from evaluation
- Agentic systems with MLflow, Agent Framework and Genie
Section 4 · Assembling and Deploying Applications22%
What the exam asks
- Code a chain using a pyfunc model with pre- and post-processing
- Control access to resources from model serving endpoints
- Code a simple chain according to requirements
- Choose the basic elements needed to create a RAG application: model flavor, embedding model, retriever, dependencies, input examples, model signature
- Register the model to Unity Catalog using MLflow
- Create and query a Vector Search index
- Identify how to serve an LLM application that leverages Foundation Model APIs
- Explain the key concepts and components of Mosaic AI Vector Search
- Identify batch inference workloads and apply ai_query() appropriately
- Configure vector search for a particular solution based on number of embeddings, update frequency, latency, and cost requirements
- Configure a persistent datastore to store and retrieve intermediate memory or structured information
- Apply CI/CD best practices such as updating a Vector Search index, promoting prompts across environments, and testing individual components of an agent
- Integrate managed, external, and custom MCP servers based on a given application requirements
- Apply prompt version control and manage prompt lifecycle
- Develop an appropriate interactive user facing interface for an agent usage scenario (Apps, Slack, Teams, etc.)
Lessons
- Chains, pyfunc models and packaging a RAG application
- Mosaic AI Vector Search: concepts, indexes and configuration
- Serving, securing and exposing LLM applications
- Operating agents: memory, MCP, prompts and CI/CD
Section 5 · Governance8%
What the exam asks
- Use masking techniques as guard rails to meet a performance objective
- Select guardrail techniques to protect against malicious user inputs to a Gen AI application
- Use legal/licensing requirements for data sources to avoid legal risk
- Recommend an alternative for problematic text mitigation in a data source feeding a GenAI application
Lessons
- Governing GenAI Apps: Masking, Malicious Input, Licensing and Problematic Text
Section 6 · Evaluation and Monitoring12%
What the exam asks
- Select an LLM choice (size and architecture) based on a set of quantitative evaluation metrics
- Select key metrics to monitor for a specific LLM deployment scenario
- Evaluate agent performance using MLflow scoring and tracing
- Use inference logging to assess deployed RAG application performance
- Use Databricks features to control LLM costs
- Use inference tables and Agent Monitoring to track a live LLM endpoint
- Identify evaluation judges that require ground truth
- Use AI Gateway (Inference Tables, Usage Tables, and rate limiting) to track an LLM or agent deployed via Agent Framework
- Use Databricks custom Scorers for evaluating agents and LLMs
- Incorporate SME feedback to improve agent performance
Lessons
- Choosing an LLM, Monitoring Metrics and Controlling Cost
- Evaluating Agents with MLflow: Tracing, Judges and Custom Scorers
- Monitoring Live Agents: Inference Tables, AI Gateway, Agent Monitoring and SME Feedback
14-Day Standard Plan
The default path for engineers who have built something with an LLM but not the full Databricks stack. It covers all 19 lessons and all 7 labs in section order, with a section drill after each section and daily flashcards. Application Development (g3, 30%) and Assembling and Deploying (g4, 22%) get the most days because together they carry more than half of the scored questions. Mock Exam 1 is on day 13 and Mock Exam 2 on day 14, so you sit the exam on day 15. Every lab runs on Databricks Free Edition except the steps marked as needing a paid workspace. About 1.5–2.5 hours a day.
- Day 1Kick-off: the exam, the platform changes and the I/O contract
- Day 2Chains, tools and Agent Bricks, then the g1 drill
- Day 3Source documents, extraction packages and chunking
- Day 4Advanced retrieval, re-ranking and the g2 lab
- Day 5Frameworks, prompt augmentation and guardrails
- Day 6Choosing the LLM and the embedding window, with the g3 lab
- Day 7Agentic systems and Genie, then the g3 drill
- … and 7 more days, with mock exams at the end
Sample questions
Written for practice in the style of the real exam: never questions from the real exam.
Sample question 1
An evaluation dataset contains realistic questions but no reference answers or expected facts yet. Which MLflow built-in retrieval judge cannot be used until ground truth is added?
- A
RetrievalRelevance - B
RetrievalGroundedness - C
RetrievalSufficiency - D
RelevanceToQuery
Show the answer
C. RetrievalSufficiency checks whether the retrieved context contains everything needed to produce the expected answer, so it needs expectations such as expected facts or an expected response. RetrievalRelevance and RetrievalGroundedness judge relevance and groundedness from the trace alone, provided it contains a RETRIEVER span.
Sample question 2
Regional sales managers spend their day in Slack and want to ask questions such as 'What were Q3 bookings in my region?' about governed sales tables. Each answer must respect the asking manager's own Unity Catalog permissions. The data team already maintains a Genie Agent (formerly Genie space) over those tables and wants the least custom code. Which approach fits?
- ABuild a Slack bot on Databricks Apps that calls the Genie Agent with the app's service principal
- BInstall the Databricks Genie app for Slack and have channel owners pin the sales Genie Agent
- CGive an agent a Slack tool backed by a Unity Catalog HTTP connection so that it can post answers in channels
- DShare the Review App of a sales agent in the Slack channel so that managers can ask their questions there
Show the answer
B. The Databricks Genie app for Slack (Public Preview) lets users DM the app or mention @Genie, and it answers with the credentials of the Slack user who asks, so Unity Catalog permissions apply per person. When a channel owner pins a Genie Agent, questions go only to that agent. It needs no custom code: an account admin enables the preview, and a Slack admin installs the app.
About this exam
What is the passing score?
Databricks doesn't publish it, because it can change. This course's readiness target is ≥ 80% on both mock exams on the first attempt, with every section drill ≥ 80%.
How many questions come from each section?
The guide gives weights, which on 45 scored items come to about 6 for Design Applications (14%), 6 for Data Preparation (14%), 14 for Application Development (30%), 10 for Assembling and Deploying (22%), 4 for Governance (8%) and 5 for Evaluation and Monitoring (12%). Each mock exam here uses exactly those counts.
Are there unscored questions?
Possibly. Up to 10 unscored items may appear. They aren't marked, they don't count, and the 90-minute limit already includes time for them, so treat every item as scored.
Does the exam have multiple-select questions?
Yes. The guide says multiple-choice and multiple-selection, and two of its ten sample questions are choose-two. The stem tells you how many to pick. About one question in seven in this course is multiple-select.
The docs call Agent Framework deployment, Knowledge Assistant and AI Gateway "legacy". Which version does the exam test?
The exam guide published on March 18, 2026. Answer with that workflow: models from code, Unity Catalog registration, agents.deploy(), Agent Bricks builders, AI Gateway on serving endpoints. The lessons flag what replaced each one so you aren't confused by current docs, and the Strategy page has the full table.
Should I learn the Agent Bricks CLI, Agent Runtime and Unity Gateway?
Recognise them, because current docs and newer questions mention them, but put your time into the exam-guide topics first. Know that the CLI is the recommended path for new agents, that Agent Runtime runs on Databricks Apps, and that Unity Gateway (GA August 4, 2026) governs model and MCP services with Beta service policies.
Will the exam say Vector Search or AI Search?
The guide says Mosaic AI Vector Search and Vector Search, so expect those names. The product is now Databricks AI Search with the databricks-ai-search package, but the concepts, index types and method names are the same, so either name points to the same answer.
Can I do every lab on Databricks Free Edition?
Almost. Free Edition gives you pay-per-token models, one AI Search endpoint with one search unit, Genie, MLflow and up to three apps. It has no provisioned throughput, no Direct Vector Access and no Knowledge Assistant, and AI Gateway inference and usage tables need a paid workspace. Lab steps that need one say so.
How much code do I need to read?
Enough to recognise the right call in a short snippet: LCEL chains, a PythonModel with load_context and predict, log_model arguments, similarity_search, ai_query, mlflow.genai.evaluate and scorers. Most questions are about which tool or design fits, not API trivia.
Do I need the Machine Learning Associate first?
No. The exam has no prerequisite. This course assumes the MLflow and model-serving basics from the ML Associate course, so skim those topics if they're new to you.
See where you stand in 8 minutes
Ten questions from this course, weighted like the real exam, with your score by section and a rough chance of passing.