Creator of Inception/GoogLeNet · Math Inc co-founder · ex-xAI co-founderCreator of Inception/GoogLeNet · Math Inc co-founder · ex-xAI co-founder
deep-learningvision
Hungarian-born computer scientist. PhD from the University of Bonn. Joined Google Research where he created the Inception architecture (GoogLeNet), winner of ILSVRC 2014, and co-introduced the study of adversarial examples in neural networks with Ian Goodfellow. Also developed Batch Normalization, a technique now used in virtually all deep networks. Left Google to co-found xAI in 2023. Departed xAI in February 2025 and joined Morph Labs as chief scientist; the team relaunched as Math Inc in September 2025, which he leads alongside CEO Jesse Michael Han, pursuing verified superintelligence through autoformalization. Its agent Gauss completed Terence Tao and Alex Kontorovich's Strong Prime Number Theorem formalization project in about three weeks, and later formalized the 24-dimensional sphere-packing proof.
Hungarian-born computer scientist. PhD from the University of Bonn. Joined Google Research where he created the Inception architecture (GoogLeNet), winner of ILSVRC 2014, and co-introduced the study of adversarial examples in neural networks with Ian Goodfellow. Also developed Batch Normalization, a technique now used in virtually all deep networks. Left Google to co-found xAI in 2023. Departed xAI in February 2025 and joined Morph Labs as chief scientist; the team relaunched as Math Inc in September 2025, which he leads alongside CEO Jesse Michael Han, pursuing verified superintelligence through autoformalization. Its agent Gauss completed Terence Tao and Alex Kontorovich's Strong Prime Number Theorem formalization project in about three weeks, and later formalized the 24-dimensional sphere-packing proof.