Roee Mayerowicz - Pizza, Agents, and the Truth: Reaching Live Data in a Multi-Agent World

Can AI really tell you where to get the best pizza in New York City? Large language models will give you an answer, but it may be outdated, biased, or just plain wrong. In this session, Roee will walk through a fun and familiar challenge: finding the city’s best pizza slice. Along the way, you’ll see why multi-agent environments powered only by LLMs often fail, and how connecting them to live, structured web data changes the game.

We’ll explore the approaches Nimble uses to deliver real-time, query-ready information at scale, and show how grounding agents in trusted web data unlocks far more than pizza recommendations. The same principles apply to pricing intelligence, market analysis, and any domain where AI needs to reason, and act, based on the real world.

Roee Mayerowicz is a Senior Researcher at Nimble, where he was one of the company’s first employees and now spearheads innovation within the research team. Roee specializes in browser agents and web technologies, focusing on how browser agents can run in a reliable way at scale. His current research centers on developing more efficient methods for acquiring online data and allowing AI to control web browsers at extremely high scale. Originally from Tel Aviv, Roee recently relocated to New York.