What you'll learn

Your generative AI prototypes work great in notebooks, but they fall apart the moment you try to share them with your team: prompts change without warning, costs spiral out of control, deployments break in production, and compliance asks questions you can’t answer. In this course, MLflow, Model Serving, and Governance for Generative AI, you’ll learn to ship reliable, observable, cost-controlled, and fully governed RAG and agent applications on Databricks at enterprise scale. First, you’ll explore professional MLflow workflows for LLMs — including prompt versioning, LLM-as-judge evaluation, and real-time tracing of failing queries. Next, you’ll discover how to deploy any RAG chain or custom agent to GPU-accelerated, autoscaling Mosaic AI Model Serving endpoints in just a few clicks. Finally, you’ll learn how to centralize all LLM access through Mosaic AI Gateway with rate limiting and fallbacks, and lock everything down with end-to-end lineage, access control, and audit in Unity Catalog. When you’re finished with this course, you’ll have the skills and knowledge of MLflow, Model Serving and Unity Catalog governance needed to confidently move generative AI applications from prototype to trusted production at any company.