← All speakers

Ramakrishnan Lokanathan

Conference affiliation: Founder · eSlams · 2026

Ramakrishnan Lokanathan founded eSlams, a public arena where people and AI models compete in games. Its open-source engine lets developers inspect individual turns and replay matches, connecting a model’s final score to the decisions that produced it.

Before eSlams, Ramakrishnan founded and sold an applied AI company. His work with eSlams uses games to examine different reasoning demands: chess exposes planning over a visible board, while poker requires decisions with information missing. Turn inspection and replay let developers investigate how a model handles those constraints; a win or loss alone leaves that explanation unresolved.

Ramakrishnan also advocates application evaluation grounded in user outcomes: tests should follow from the result a system promises its user. He recommends retaining examples of earlier successes and turning production failures into regression tests. Together, these practices make previously successful behavior and newly discovered problems part of the next release’s evaluation. His talk If Your AI Scores 100%, Your Eval Is Wrong addresses this work on evaluation.

1 conference talk

Loading more talks and highlights…

References