Giving Data value to AI

Modern ML and AI systems are great at generating information — but they’re surprisingly bad at understanding what enterprise data actually means. Business logic, undocumented assumptions, and tribal knowledge are scattered across one-off SQL and people’s heads — and often never make it into models, agents, or applications at all. This talk dives into Engineering Meaning: a new systems approach for turning business intent into a first-class, versioned, testable asset — and delivering that meaning in context to AI systems, applications, and analytics. You’ll see how teams: Encode business intent into a semantic model as executable meaning Deliver definitions, relationships, and intent dynamically to agents and applications Treat context as an engineering problem with observable retrieval and evaluation Use ML-style feedback loops to continuously improve how meaning is retrieved and applied Rather than patching prompts, Engineering Meaning moves understanding into a shared system that can be versioned, governed, evaluated, and optimized. For ML, platform, and data engineers building real production AI systems, this talk frames shared understanding as a missing layer in modern architectures — and shows how treating meaning as infrastructure leads to more reliable AI, faster iteration, and fewer silent failures. Schedule: 6:15-6:30 food 6:30 pm announcements 6:45-7:30 speaker 7:30-8:30 announcements