Ibrahim Abdu is Co-Founder at Fabraix, a Y Combinator-backed AI security company building the frontier red-teaming agent that identifies failure modes and security exploits in AI systems before users encounter them.

Ibrahim holds a BA in Philosophy, Politics and Economics from Oxford University (ranked top 8% of his cohort) and is a UCL PhD drop-out, bringing rare academic depth to the applied AI security space.

A published researcher with multiple papers in Q1 journals, Ibrahim also co-authored the ACE benchmark, a dynamic metric measuring the adversarial token cost to exploit AI agents, pushing the frontier of offensive AI security evaluation.

Prior to Fabraix, Ibrahim was the first data scientist at Sequoia-backed Two, where he built a fraud engine from scratch processing over $1B in annual B2B transactions and averting $50M in losses for multiple Fortune 150 companies. He also worked as a Software Engineer at Meta building AI agents to debug and fix production errors, and built a proprietary compiler at TradingHub for algorithmic insider trading detection.

Fabraix's product Nyx has already uncovered security vulnerabilities in AI agents at dozens of Fortune 500 companies, and on AgentHarm, the leading offensive AI security benchmark, Nyx achieved a 78% attack success rate, surpassing GPT-5.6 Sol at 67%.