Databricks Specialist Sessions

From Prototype to Production with MLflow for GenAI

Duration: 2 hours

This session introduces MLflow as the open platform for building, evaluating, and deploying production-ready GenAI agents at speed, with core capabilities such as tracing, evaluation, versioning, and the AI Gateway, which provides deep visibility into agent behaviour while simplifying debugging and improving reliability at scale. 

We’ll also cover MLflow’s newest features brought to life through a live demo showcasing end-to-end tracing, evaluation, and monitoring in action. Whether you’re just starting with GenAI or scaling existing agents, you’ll leave with a clear phased adoption strategy—from initial tracing through full production monitoring—to deliver high-quality, cost-efficient, and accountable GenAI applications.

Key takeaways:

• - Use MLflow to support the full GenAI agent lifecycle, from experimentation to production monitoring.
• - Apply tracing, evaluation, and LLM judges to understand and improve complex agent behaviour
• - Leverage the newest MLflow features, including prompt optimisation and the Prompt Registry API for governed, reusable prompt management
• - Follow a phased adoption strategy for robust monitoring, evaluation, and governance of GenAI systems on Databricks

Ideal for: ML engineers, data scientists, and platform teams building or operating GenAI systems on Databricks