LLM Evaluation Summit

Evals are the only thing standing between a working AI product and a slow-moving disaster. This hybrid summit goes deep on the practices the model teams behind production AI actually use. Sessions cover: building golden datasets that survive a year of model updates, calibrating LLM-as-judge against human raters, eval CI that catches regressions before deploy, multi-turn eval design, and cost / latency budgets that don't get blown by the next model release. Attend in-person in Boston for the workshops, or stream the main track online — both ticket types include the recordings.