"Tell me about a time you had to solve a difficult generative AI problem where the root cause was unclear at first—for example poor retrieval quality, hallucinations, or unstable evaluation results. How did you decide what to investigate first, align others around your approach, and what was the outcome?"
