Building the Infrastructure of AI Trust

Anastasios Angelopoulos completed his undergraduate degree in Electrical Engineering at Stanford, then earned his PhD in EECS at UC Berkeley, advised by Michael I. Jordan and Jitendra Malik, two of the most cited names in machine learning. There, he also became a Postdoctoral Scholar under Ion Stoica who would later join him as Arena's co-founder and chairman. His research focused on conformal prediction, uncertainty quantification, and statistical guarantees for black-box systems, the math that makes it possible to say precisely what an AI model can and can't be relied on to do.

In 2023, that research became Arena. What started as a side project during his PhD — anonymize two models' outputs, let users vote, aggregate the results — is now the AI industry's most widely used model evaluation platform. Its Battle Mode generates continuous preference rankings used by OpenAI, Google, Anthropic, Meta, and DeepSeek, both before they release models and after, to monitor how they're performing in the wild. Arena's leaderboard is built from more than 82 million user votes across more than 700 million conversations, with over 10 million people visiting the platform every month from more than 150 countries.

Arena crossed $100 million in annualized revenue within eight months of launching its first commercial product, AI Evaluations, one of the fastest paths to that milestone in the industry's history. The company has raised $250 million to date, including a $150 million Series A in January 2026 at a $1.7 billion valuation.

The platform is now extending the same approach to AI agents. Agent Mode evaluates how autonomous systems perform on real tasks - building full applications, running financial models, executing multi-step workflows - as AI shifts from answering questions to taking action on its own.

Selected to join the global Endeavor Network at Endeavor's International Selection Panel in San Francisco in May 2026, Anastasios brings deep technical rigor and a research career built around a single, increasingly urgent question: as AI systems get more capable, how do we know which ones to trust.