Building Pipelines End-to-End with AI

​You build a pipeline. It breaks. You debug, fix, redeploy. Repeat.

What if an agent could handle that loop and show you the results visually so you can focus on the decision, not the plumbing?

Hugo Lu (Orchestra) and Jacob Matson (MotherDuck) will live-demo an AI agent that builds a data pipeline from scratch: orchestration in Orchestra, data processing in MotherDuck, and visual output through Dives - all wired together with MCP.

They'll show how source-controlled Dives turn pipeline outputs into reproducible, de-buggable artifacts that help teams make decisions faster.

You'll see:
• An agent scaffold and run a pipeline from a natural language prompt
• MCP connecting orchestration, warehouse, and visualization layers
• Visual debugging of pipeline results with MotherDuck Dives
• "Blessed Dives" - versioned visualizations that live in source control

Live Q&A at the end - bring your gnarliest pipeline problems.