Building data agents enterprise can actually trust

Building data agents enterprise can actually trust Your data agent is confident. Is it right? Everyone is building agents. Far fewer are building data agents that can reliably reason over real enterprise data — with all its messiness, missing context, and scale — and give you a way to know when they're wrong before your stakeholders do. The hardest part of a data agent isn't the LLM. It's everything underneath: the semantic layer, the ontology, the data federation logic, and the infrastructure to serve it reliably. Get that wrong and your agent returns confident wrong answers. And because it sounds authoritative, those wrong answers are more dangerous than no answers at all. Join us for an evening with Jerry Xu, AI Infrastructure Architect at Oracle, who has spent his career building the systems that put models into production at Meta, Lyft, Box, and the company he founded, DataTron. This is a working session for technical leaders, founders, and data/AI platform teams who are past the demo stage and now have to make data agents dependable. What Jerry will cover: • The semantic layer is the real challenge — why NL2SQL alone breaks on real enterprise schemas, and why ontologies, federation logic, and certified data context are where the hard engineering actually lives • Building the data agent stack — agent architecture, orchestration patterns, multi-agent coordination, and how to handle ambiguity when querying live, messy enterprise data • How do you know when to trust the answer? — why standard benchmarks miss the failure modes that matter in production, what a real grading pipeline looks like, and how to design for verifiability from day one • What breaks at scale — failure modes no one talks about until it's already happened, and why long-term reliability requires treating models as living systems Who should come: Engineering and AI leaders, data platform and infrastructure teams, founders building data products, and operators responsible for making AI reliable in production.Hosts includeOpen Future ForumHG insightsAgentic Fabriq More on Jerry Xu Jerry is an AI platform architect focused on generative AI and ML infrastructure, with a background spanning Meta, Twitter, Microsoft, and a company he founded himself. His through-line is the unglamorous but decisive work of making AI systems run in production. Career highlights: His strength sits at the intersection of AI research and production-grade infrastructure: distributed systems, model lifecycle, MLOps, and the identity and security questions that come up as agents start acting autonomously. That last point pairs naturally with Agentic Fabriq's focus, which makes him a strong fit for this room.