Mehtaab Sawhney

Mehtaab Sawhney (born September 20, 1998) is an American mathematician, specializing in combinatorics and number theory. He is an assistant professor at Columbia University.

Early life and education Sawhney grew up in Commack, New York. While attending Commack High School, Sawhney competed in the United States of America Mathematical Olympiad (USAMO), and participated in MIT PRIMES. In 2016, for his research in graph theory through MIT PRIMES, he was named a semifinalist in the Intel Science Talent Search and won an award at the International Science and Engineering Fair.

Sawhney attended the University of Pennsylvania for one year before transferring to the Massachusetts Institute of Technology (MIT), where he studied mathematics and computer science. and won the Morgan Prize in 2021. He completed his PhD in 2024, supervised by Yufei Zhao.

Career At MIT, Sawhney was noted for his productive collaboration with fellow student Ashwin Sah in the fields of combinatorics and number theory. Sah and Sawhney each attended MIT for undergraduate and doctoral studies between 2017 and 2024. During that time, they co-authored more than 50 papers, on topics such as Ramsey theory, Steiner systems, subspace designs, and Szemerédi's theorem.

Sawhney was awarded a Clay Research Fellowship in 2024, He became a tenure-track assistant professor of mathematics at Columbia University in 2024. He went on leave from Columbia in 2026 to join the artificial intelligence company OpenAI as a mathematician.

In 2025, Sawhney and Mark Sellke used GPT-5, a large language model developed by OpenAI, to search the literature on several conjectures made by Paul Erdős in combinatorial number theory. In 2026, Sawhney and Sellke used GPT-5 to find a counterexample to Erdős's conjecture on unit distance graphs. GPT-5 provided a method to construct n points in the plane from which Ω(n1+ε) pairs of points are unit distance apart, for some ε > 0. A team of mathematicians simplified the argument and computed an explicit lower bound for the exponent ε to be 6 × 10−38; Will Sawin subsequently improved the exponent to ε ≈ 0.014.