Databricks Foundational Workshops

Take a hands-on journey through building and hardening a production-grade AI agent on Databricks. You'll deploy a custom customer-support agent as a Databricks App, govern it with your own identity, confront the data quality challenges that trip up real agents, then use LLM judges to measure quality, fix the root cause, and prove the fix with numbers. Instead of hand-writing code, you'll drive Genie Code, an in-workspace coding agent, using plain English.

In 90 minutes, you'll walk away knowing how to build an agent from a model, tools, and memory, keep governance flowing through the app on behalf of each user, and turn "it feels better" into a measurable quality score.

You'll learn how to:

• Deploy a custom agent as a Databricks App from a single plain-English prompt using Genie Code

• Give the agent tools with Unity Catalog functions and Vector Search over governed company data

• Persist every conversation to Lakebase, managed Postgres running next to the Lakehouse

• Enforce governance with on-behalf-of-user access, so Unity Catalog column masks redact PII through the deployed agent with zero extra code

• Understand how data quality drives agent behaviour, the most common failure mode in enterprise agents

• Measure response quality with MLflow LLM judges, fix the root cause, and watch the score flip

• Productionize the loop by packaging the app and its evaluations as a Declarative Automation Bundle, with traces governed and queryable in Unity Catalog

Who should attend
If you're a data engineer, ML practitioner or technical architect who wants hands-on experience building, governing, and evaluating AI agents on the Databricks Data + AI Platform, this session is for you. No prior agent development experience required, and all lab environments are pre-configured.

Agenda

• 5 mins — Welcome

• 15 mins — Presentation: From Prompt to Production, Building and Hardening AI Agents

• 1 hour — Hands-on Lab

  • Part 1: Build and deploy your agent as a Databricks App

  • Part 2: Govern it with your own identity (on-behalf-of access and column masking)

  • Part 3: Understand agent quality and measure it with LLM judges

  • Part 4: Fix the root cause, prove it with metrics, and productionize

• 10 mins — Q&A 

   

This introductory, hands-on workshop is built for data engineers and data architects who want practical experience with modern data engineering on Databricks.

You’ll work directly in your own lab environment and learn how to design, build, and operate reliable data pipelines using real-world use cases. You’ll leave with hands-on experience using GitHub, ingesting data from multiple sources, and operationalizing pipelines end to end.

In this session, you will learn:

• Introduction to Lakeflow and the Data + AI Platform

• Getting started with Declarative Pipelines, Streaming Tables, and Materialized Views

• Working with a realistic data streaming use case

• Mastering Lakeflow Jobs with advanced control flow and triggers

• Generative AI for Data Engineers: Genie and Databricks Assistant

• Understanding data governance and lineage with Unity Catalog

• Benefits of Serverless Compute

Effective governance is essential to democratizing data and AI in the age of LLMs and generative AI. Enterprises struggle with fragmented governance across databases, warehouses, lakes, and diverse data formats, which complicates discovery, access permissions, data usage, audits, and sharing. Additionally, users often cannot access the right data from their preferred tools, leading to vendor lock-in. Managing ML models and AI tools for regulatory compliance adds another layer of complexity.

Learn how Unity Gateway helps organizations lower AI costs without productivity. In this 90-minutes hands-on workshop, you will configure and use a governed Model Service, add metadata for usage analysis, query cost and audit evidence, configure rate limits and budgets, manage Unity Catalog permissions, and test runtime service policies on model requests and responses.

In this session, you will learn how to:

• Configure a governed Model Service for AI workloads

• Apply rate limits and budgets to manage AI costs

• Enforce guardrails and service policies on requests and responses

• Governed access to MCP

• Unity Gateway for Coding Agents

• Use observability and audit evidence to analyze usage, cost, and policy outcomes

Agenda:

• Introduction to Unity Gateway

• Unity Gateway Deep Dive & Hands-on Lab

• Q&A and Wrap-up

Most organizations run complex cloud data architectures that silo applications, users and data. As a result, most analysis is performed on stale data and there’s no single source of truth of data for analytics - making it difficult to scale trusted insights across teams.

Join this interactive hands-on workshop to learn how Databricks Lakehouse enables a multicloud lakehouse architecture that delivers data warehouse performance at data lake economics — with up to 12x better price/performance than traditional cloud data warehouses. You’ll discover how data architects can unify infrastructure and governance, while data analysts and BI teams can work with the freshest, most complete data—and accelerate insights using Databricks’ native AI/BI capabilities along with their favorite BI tools.

Whether you’re building a scalable foundation or delivering insights to the business, this workshop is designed to help you modernize your analytics stack and make AI-powered BI a reality.

In this workshop, you’ll learn:

• How the Data + AI Platform and Databricks Lakehouse support a unified, multicloud lakehouse architecture
• How to manage and monitor compute, data access, and users across your lakehouse using Databricks Lakehouse
• How data teams and analysts can query directly on your data lake using the built-in SQL editor, AI/BI or their preferred BI tools.
• How to use AI-assisted features like AI-assisted code completion, automated dashboard creation and natural language querying via AI/BI Genie,to increase productivity and scale self-service analytics.

Most application stacks are fragmented. Transactions live in one system, analytics in another, and pipelines try to stitch everything together.

That architecture slows development. Teams often spend more time debugging data movement than building features.

This workshop introduces a different approach. You will work with Lakebase, a fully managed Postgres database on Databricks that sits alongside the lakehouse, allowing your applications, analytics, and AI to operate on the same data.

The session demonstrates how modern application teams can move faster when apps, analytics, and AI run on the same data. You will work with Lakebase to spin up a database, connect with Python, run queries, and power real application workflows, showing how the lakehouse can support production-ready applications.

What You'll Learn:

• Run your application and analytics on the same data
• Set up Lakebase on Databricks and start using it quickly
• Connect your application using SQL, Python, and standard tools
• Build features on live data instead of batch copies
• Push lakehouse data directly into your application
• Apply governance across both data and applications
• Get closer to something you could run in production