Higher School of Mathematics webinar tackles LLM evaluation for production teams
Free May 28 session moves teams from ad-hoc prompt testing to repeatable metrics, with live tooling demo and take-home framework.
A free online workshop on May 28, 2026 at 19:30 Moscow time aims to shift LLM feature development away from the familiar "poke the prompt until it works" pattern. Hosted by the Higher School of Mathematics, the session promises a live end-to-end evaluation cycle — from raw log collection through automated quality measurement — along with a take-home framework teams can deploy immediately.
Andrey Kiselev, Head of Product at an AI company and formerly at Revolut and Yandex, and Fedor Azarov, who leads data research at Sber CIB, will walk through systematic approaches to improving AI products. The agenda includes translating user feedback into quantifiable metrics, measuring the economic impact of new features, and setting up repeatable evaluation pipelines. Product managers looking to escape anecdotal success stories and engineering leads trying to standardize AI workflows in their teams are the stated audience.
The organizers emphasize that the webinar will go beyond methodology slides. Attendees will see a real-time demonstration of the full evaluation stack — tooling choices, automation setup, and metrics dashboards — rather than abstract best practices. The session is open to commercial and side-project builders alike, with registration available through the Higher School of Mathematics. Whether the framework proves portable across different model families and whether the tooling recommendations favor open-weight or API-based workflows will become clear once the session airs.


