Paul Lezeau is a researcher and maths PhD candidate at the London School of Geometry and Number Theory, currently working at Google DeepMind where he bridges rigorous mathematics and AI research. His work spans algebraic geometry, number theory and the mathematics of AI, with applied projects on robustness of deep learning and formal verification for theorem proving. He has formalized nontrivial algebraic number theory results in Lean and contributed to mathlib3, helping OpenAI and the community test neural theorem provers with a curated database of formalized problems. Paul combines strong academic training from Imperial, Cambridge and Warwick with hands-on industry experience (AlphaProof, DeepMind), and a knack for translating deep theory into reproducible code and tutorials. An interesting quirk: alongside abstract algebraic work he has published practical adversarial-ML tutorials and exam-marking experience, reflecting a rare mix of formal, applied and educational contributions.
5 years of coding experience
Doctor of Philosophy - PhD Mathematics , Doctor of Philosophy - PhD Mathematics at Imperial College London
Master's degree Mathematics, Master's degree Mathematics at University of Cambridge
BSc Mathematics, BSc Mathematics at University of Warwick
Visiting Student Theoretical and Mathematical Physics, Visiting Student Theoretical and Mathematical Physics at Shanghai Jiao Tong University
Lean 3's obsolete mathematical components library: please use mathlib4
Role in this project:
Backend Developer
Contributions:90 reviews, 310 commits, 32 PRs in 1 year 5 months
Contributions summary:Paul primarily contributed to the mathematical components library by adding new lemmas and theorems. They focused on extending the library's functionality, adding results on topics such as quotients, ideals, and unique factorization domains. Furthermore, they collaborated with others on adding and refining theorems within the ring theory and algebraic structures portions of the library. These contributions improved the formalization of mathematical concepts within the Lean 3 framework.
Repository for formalization of the Polynomial Freiman Ruzsa conjecture (and related results)
Contributions:30 pushes, 7 branches in 11 months
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.