3 Engineers Reveal What's Broken in AI & How to Fix It | SF AI Engineers Meetup Feb 2026

This SF AI Engineers meetup will change how you think about building AI systems. No vendor pitches, no hype cycles, just three engineers sharing what they actually learned after years of building production AI. Scott Clark (Distributional CEO) spent 10 years optimizing AI performance before realizing he was solving the wrong problem. It's not performance that's killing AI adoption; it's confidence. His pivot from testing to analytics treats agents like users in web apps, tracking their journeys through tool graphs to find patterns you didn't know existed. YK (Daft Engineering) breaks down why most engineers are stuck at Level 1 of AI coding when there are actually 4 levels to master. From "letting AI go wild" (fine for prototypes, terrible for production) to staff engineer-quality code that requires line-by-line understanding of every generated line. Wassim (Tesla Energy) had a simple insight: engineers don't diagnose problems by reading CSV files. They look at charts. So he taught LLMs to do the same by feeding them screenshots instead of raw time series data. Vision-based diagnosis for technical systems.