Lechao Xiao
Research Scientist at DeepMind
Champaign, Illinois, United States
Summary
🤩
Rockstar🎓
Top SchoolLechao Xiao is a research scientist with six years of experience at the intersection of mathematics and machine learning, currently working at DeepMind after several research roles within Google Brain. He holds a PhD in Mathematics from UIUC and brings deep theoretical insight to practical ML problems, contributing to high-profile open-source projects such as google/neural-tangents (adding optimizers, activations, and input-product primitives). His career blends academic rigor—formerly a Hans Rademacher instructor—with hands-on engineering, enabling robust implementations of infinite-neural-network tools. Based in Champaign, Illinois, he is comfortable moving between theory and code, translating mathematical structures into scalable model components.
6 years of coding experience
4 years of employment as a software developer
Doctor of Philosophy - PhD Mathematics, Doctor of Philosophy - PhD Mathematics at University of Illinois Urbana-ChampaignBachelor of Science - BS Mathematics, Bachelor of Science - BS Mathematics at Zhejiang University
Github Skills (7)
neural-network10
mle10
gradient-descent10
jax10
python10
ml10
machine-learning9
Programming languages (2)
Github contributions (5)
google/neural-tangents
Jun 2019 - Dec 2022
Role in this project:
ML Engineer
ML EngineerContributions:1 review, 17 commits, 13 comments in 3 years 6 months
Contributions summary:Lechao contributed several enhancements to the `neural-tangents` repository, a project focused on infinite neural networks. They implemented the `max_learning_rate` function for momentum optimizers, which is critical for model training. The user also added the `Gelu` activation function and the `Sign` function. Further, they implemented the `FanInProd` which enables the Hadamard product of inputs.
kernelpythonbayesian-inferencebayesian-networksgaussian-processes
SiuMath/jax
Dec 2019 - Dec 2019
Composable transformations of Python+NumPy programs: differentiate, vectorize, JIT to GPU/TPU, and more
Contributions:8 pushes, 2 branches in 6 days
pytorchpythonjitgpunumpy
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.