Lessons from medical QA for AI metrics

Sarah Gebauer is a physician and founder of Validara Health who works at the intersection of clinical AI evaluation and governance. We cover what it actually means to evaluate an AI product for clinical safety, how physicians can draw on skills they already have to work effectively with these systems, and where those analogies break down. The core tension running through the conversation is that we are layering systems with unknown failure modes on top of systems with known ones, and hoping the combination is better, but that's a hypothesis, not a guarantee.