Valeria Zhang

As analytics and AI workloads grow, most enterprises scale by adding more Databricks workspaces and infrastructure. Over time, this leads to admin overhead, duplicated standards, slow onboarding, and analytics, ML and GenAI teams operating in silos. In this session, we share our lessons learned at Procter & Gamble, one of the largest Databricks workspace deployments. We show how we reversed that pattern by designing a multi-tenant Databricks architecture built to adapt to constant change and scale with minimal admin effort. Databricks is a multi-tenant platform — this session shows how to maximize it at enterprise scale. Using a “hub-and-spoke” approach, we consolidated workspaces while preserving team autonomy and execution speed. Attendees will learn how to: - Design a multi-tenant architecture that scales without increasing admin overhead - Consolidate workspaces while maintaining team ownership - Run analytics and AI workloads on a shared platform without creating silos