Full talk · edited transcript

User-Guided LLM Answer Quality Evaluation

Ivan Bushmarinov23 May 2026

Ivan Bushmarinov, Engineering Manager, Answer Quality, Perplexity.

The talk is about doing analytics at scale when there is not enough manpower to validate every case in which an AI gives a verdict. Perplexity is a reasonably small startup: there are barely enough people to code features and never enough to validate them. The talk covers lessons from deriving an online quality signal from user behavior, and from building an internal platform for offline evaluation.

Continue reading — it's free

Google or a one-time email code. No password.