Databricks at AWS
re:Invent

More than 80% of AI projects fail. The reason? Customers aren’t ready to put AI in production in areas where the agent can cause reputational or financial harm. Agent Bricks solves this by automatically generating domain‑specific synthetic data and task‑aware benchmarks, then using those benchmarks to optimize the agents for accuracy, latency and expense — eliminating the tedious trial‑and‑error cycle. In this session, you’ll see how to design and scale multi‑agent systems that combine retrieval, reasoning and generation to answer complex, enterprise‑grounded queries, while tracking experiments, fine‑tuning models and building governed, repeatable workflows that can be deployed at scale on AWS.

Legacy data warehouses often slow innovation with outdated code, silos and costly migrations. In this session, we’ll share proven best practices to modernize with confidence. Learn how to profile and assess source systems, convert legacy SQL at scale with automation and validate data integrity to avoid surprises. We’ll cover an end-to-end process — from discovery through reconciliation — that reduces risk, improves predictability and accelerates migration timelines by up to 50%. Whether you’re moving one system or many, you’ll leave with practical tactics to streamline your journey to a modern lakehouse.