From CERN to the Enterprise: Building a Trusted Data Foundation Across 100+ Source Systems and Why AI Agents Depend on It

What do particle physics experiments and enterprise AI platforms have in common? One probes the fundamental laws of the universe; the other powers everyday business decisions across the enterprise. Both live or die on the same thing: turning enormous volumes of raw data into something people and increasingly, autonomous agents can actually trust. Having worked with petabytes of scientific data at CERN and later architected a cloud-based enterprise data platform integrating more than 100 source systems, I've found the core principles remarkably stable even as the technology churns. Lineage, reproducibility, governance and quality were non-negotiable in physics long before anyone called them "AI-ready." This session walks through that architecture in practical detail: metadata-driven ingestion that scales without a new pipeline per source; reusable integration patterns across operational, transactional, and financial systems; data contracts and quality scoring that catch drift before it reaches consumers; end-to-end lineage from source system to semantic layer; and the self-service model that keeps a central team from becoming the bottleneck. The last part is where it gets interesting. Humans cope with bad data, they ask a colleague, they notice a number looks wrong, they apply context no system ever captured. Agents don't. An LLM querying an undocumented table with no quality signal will produce a confident, plausible, wrong answer and act on it. Everything that was "nice to have" in the dashboard era becomes load-bearing infrastructure the moment agents start consuming your data directly. Attendees will leave with concrete architectural patterns, honest lessons from what didn't work, and a clear view of which foundational investments actually pay off when the AI layer arrives whether they're modernizing an existing platform or starting from scratch. Because before an organization can build intelligent systems, it has to build a data platform those systems can trust.
Format: TECHNICAL TALK
Location: ROOM B
Speakers: Gordana Milutinovic Dumbelovic
Level: Beginner to Intermediate
Audience: Registered participants