Databricks GenAI Mock Exam: 8 Questions
Test practical knowledge of RAG, data preparation, deployment, governance, and evaluation using the supplied Databricks mock exam material.
Try this quiz
Play through the questions and see your score instantly
Ready to test your knowledge?
8 questions · Quick play · Instant results
Make your own quiz videos
Turn any topic into a polished video quiz — with AI-powered questions, voiceover, and animations. No video editing skills needed.
Unlimited quizzes, free to start
Create as many quizzes as you want. Describe your topic and AI builds the questions, answers, and explanations for you.
Customise everything
Pick from stunning templates, tweak colours and fonts, add your branding, and choose between vertical or landscape formats.
Export-ready videos
Download HD videos optimised for TikTok, YouTube Shorts, Instagram Reels, or full-length YouTube — one click, no editing.
No credit card required
Quiz Questions & Answers
Review every prompt, the correct responses, and helpful context to prep for your own run-through.
Question 1: A RAG chain must fetch relevant passages before generating an answer. Which sequence matches the recommended flow?
Output parser, retriever, context prompt, LLM
Retriever, context prompt, LLM, output parser
LLM, retriever, output parser, context prompt
Context prompt, output parser, LLM, retriever
Question 2: Procedures in a Markdown manual are split across fixed-size chunks. What adjustment best preserves their structure?
Split by headings and sections, with a size cap and small overlap
Increase the generation model’s temperature
Remove the document headings
Use even smaller fixed character windows
Question 3: A team finds many invented citations while reviewing sample answers. What is the best next step?
Increase the model’s maximum output tokens
Wait until all evaluation metrics are finalized
Turn the pattern into tests and scorers, then improve grounding
Delete the reviewed responses
Question 4: A deployed agent fails with permission errors even though it worked in a notebook. What should the team check?
Whether the model temperature is too low
Whether required resources were declared and endpoint privileges granted
Whether the vector index should be made public
Whether document chunks need more overlap
Question 5: A 40-million-vector index needs low latency, about 150 queries per second, and updates visible within seconds. Which setup fits best?
A separate in-memory index on every application instance
A standard endpoint with continuous sync and added capacity if needed
A local index rebuilt manually every hour
A storage-optimized endpoint with triggered sync
Question 6: A team wants to register an MLflow-logged agent for governance. Where should it be registered?
As files emailed to the platform team
As a local pickle file
In Unity Catalog with a three-level name
In the DBFS root directory
Question 7: Which statement best distinguishes offline evaluation from production monitoring?
Both activities happen only after release
Monitoring removes the need for pre-release evaluation
Evaluation tests versions on a fixed dataset before release; monitoring scores live traffic
Evaluation measures only cost, while monitoring measures only accuracy
Question 8: A chatbot must always return valid JSON with required keys. Which scorer is the best fit?
A code-based custom scorer
A Guidelines judge only
A Safety judge
A RetrievalGroundedness judge