Generative AI on Databricks¶
Welcome. Over two days you will build five working AI agents on Databricks without writing more than a line of SQL, then pitch a use case of your own.
| Day | Lab | What you build |
|---|---|---|
| 1 | Lab 1 | An LLM chat with a tool attached, in AI Playground |
| 1 | Lab 2 | A Genie Agent that answers questions about work orders |
| 1 | Governance, evaluation and cost | How you'll judge and control what you build |
| 1 | Lab 3 | AI over IT tickets with one-line SQL functions |
| 1 | Lab 3b | Two Genie Code skills: one captured for you, one written by hand |
| 2 | Lab 4 | A Knowledge Assistant that answers from public regulations |
| 2 | Extra: RAG by hand | Optional: extract, chunk, embed and search the regulations yourself in SQL |
| 2 | Lab 5 | A Supervisor Agent that routes between data and documents |
| 2 | Capstone | Your own use case on one page, pitched to a manager |
Data rule for both days
Only public documents and synthetic datasets are used in these labs. Never upload internal, operational, plant or personal data into any workspace during the training.
How the labs work¶
- Each lab has Steps (do these), Check yourself (questions to confirm you understood), and a Stretch task for when you finish early.
- Work in pairs. One drives, one reads the steps aloud, swap halfway.
- When something breaks, raise a hand. Broken things are the most useful part of a lab.
Start with Before you start.