The Agentic Lakehouse: Optiver’s AI Data Stack

This session deep-dives into how we’ve integrated AI into the core of our data workflows.We will showcase three key innovations:

• Custom Databricks MCP: How to deploy a centralized MCP deployment enabling LLMs to interact seamlessly with lakehouse resources while maintaining strict context management.
• Self-healing dashboards: A “prompt-to-production” workflow that automates dashboard deployment and utilizes Playwright-based agents to autonomously detect and fix UI or deployment errors.
• Agentic research: Moving from static answers to reproducible insights. We’ll demonstrate agents that conduct deep-dive research and generate both the results and the re-runnable Databricks notebooks required for human validation.