The Data Layer Underneath Every AI Agent: Databricks' Co-Founder on Scaling to $5B+ ARR

Everyone is talking about AI agents. Almost nobody is talking about what's underneath them. Every agent, every model, every workflow is only as good as the data layer it runs on. Databricks has spent over a decade building that layer, and it's now a $5B+ ARR company valued at $134B, growing 65% year over year with an IPO on the horizon. Arsalan co-founded Databricks out of the Apache Spark project at Berkeley. He's spent 12 years watching enterprises try to get their data right, first for analytics, then for ML, now for AI agents. Most of them are still getting it wrong. In this session, Arsalan will cover: Why the companies winning with AI agents all have one thing in common: they fixed their data layer first The shift from massive foundation models to bespoke, domain-specific models built on enterprise data, and why that shift changes everything about how you build What "data quality" actually means when your agents are making decisions, sending emails, and talking to customers autonomously How Databricks scaled from open-source project to $5B+ in revenue, and the go-to-market lessons that came with growing 65% at that scale Why he believes the revenue pyramid is about to invert: in five years, most of the value will sit in AI applications, not infrastructure The mistakes he sees founders make when they skip the data foundation and go straight to building agents