Keller Jordan
Keller Jordan is a researcher and engineer at OpenAI, which he joined in December 2024. He is best known for creating the Muon optimizer, a neural network training algorithm for hidden layers that uses Newton-Schulz iterations to orthogonalize momentum-based updates, demonstrating substantially faster convergence than standard optimizers. He holds bachelor's degrees in mathematics and computer science from UC San Diego and previously worked as a machine learning engineer at Hive.
“The Muon optimizer's developer, Koiler Jordan, was recruited by OpenAI in December 2024 precisely because of this work — he was originally an independent developer.”
Source→“The Muon optimizer's developer, Koiler Jordan, was recruited by OpenAI in December 2024 precisely because of this work — he was originally an independent developer.”
Source→AI-extracted from podcast / newsletter / paper summaries. May contain errors.