This isn't another "watch and forget" course — it's a hands-on, build-first program where you create four real, production-style AI agent systems on the Databricks platform. Each project mirrors challenges that real companies solve today, using the same tools, patterns, and architecture used in actual industry deployments: Mosaic AI, Unity Catalog, Vector Search, LangGraph, MCP servers, and Databricks Apps. By the end, you won't just understand these tools conceptually — you'll have shipped four complete, portfolio-ready projects you can demo, extend, and talk about in interviews.

1. Data Dictionary App on Unity Catalog

Most data teams struggle with undocumented, hard-to-navigate schemas — this project solves that problem for good. You'll build a production-grade Databricks App paired with an MCP server that automatically reads Unity Catalog schemas, understands table and column structures, and uses LLMs to generate detailed, human-readable documentation for every column. All of this is then served through a natural, conversational chat interface so anyone on the team — technical or not — can ask questions about the data and get instant, accurate answers. You'll learn how to work with Unity Catalog metadata programmatically, design an MCP server from scratch, and connect LLM-generated content to a live application layer.

2. Resume and JD Matcher System

Recruiting teams spend hours manually screening resumes against job descriptions — this project automates that process end to end. You'll use LLMs to parse and structure raw resume data, then embed both resumes and job descriptions into Databricks Vector Search for fast, semantic retrieval. On top of that, you'll build a LangGraph agent that uses UC Functions to intelligently score, rank, and explain why a candidate is (or isn't) a strong fit for a role. This project teaches you the full RAG pipeline — from embeddings and vector indexing to agentic reasoning and explainability — in a real recruiting-tech context.

3. Hospital Agent

Healthcare workflows are high-stakes, tightly regulated, and full of edge cases — which makes this one of the most demanding and valuable builds in the course. You'll create a prior-authorization agent using a Streamlit-based Databricks App, an MCP server, and LangGraph orchestration to manage multi-step decision-making. The agent pulls live patient context directly from Unity Catalog tables, cross-references clinical guidelines, drafts authorization approvals automatically, and knows when to escalate complex cases to a human reviewer. This project pushes you to think about reliability, compliance, and safe automation in a regulated industry — skills that translate directly to enterprise AI work.

4. Support Ticket Triage

Customer support teams are often overwhelmed with incoming tickets that need to be sorted, prioritized, and routed quickly and accurately. In this project, you'll build an agent that classifies tickets, assigns priority levels, and routes them to the correct team based on historical resolution patterns — then automatically drafts a first-response message to the customer. You'll also implement MLflow evaluation to continuously measure agent performance, add a human-in-the-loop review step for quality control, and deploy the final agent using Mosaic AI serving. This is a complete, production-grade workflow from model evaluation to live deployment.

Across these four projects, you'll gain practical, transferable experience in agent orchestration, retrieval-augmented generation, tool calling, vector search, governance, model evaluation, and deployment — all within the Databricks ecosystem. Whether you're aiming to break into AI engineering, upskill in your current data role, or build a strong project portfolio, this course gives you real, defensible work to show for it.