Huamin Li
Machine Learning Software Engineer at Facebook
San Francisco Bay Area United States
Summary
🤩
Rockstar🎓
Top SchoolHuamin Li is a Machine Learning Software Engineer based in the San Francisco Bay Area with a decade of experience building AI infrastructure and production ML systems. At Facebook/Meta he contributes to the developer platform for Caffe2, PyTorch and ONNX, focusing on compiler and backend work that helps bridge research models to production—work visible in notable open-source projects like PyTorch and Glow. His contributions span AOTInductor compilation, quantized operators, NNPI test hardening and automation, demonstrating a mix of low-level systems engineering and ML model deployment expertise. Prior roles include recommender/search work at Roku and quantitative research at JPMorgan, reflecting a comfort with both applied ML and rigorous numerical methods. He holds an applied mathematics PhD from Yale and often combines academic rigor with pragmatic engineering, such as adding targeted test flags and pragmatic workarounds to keep large-scale ML toolchains robust.
10 years of coding experience
1 year of employment as a software developer
De Anza CollegeBachelor of Arts (B.A.), Applied Mathematics, Bachelor of Arts (B.A.), Applied Mathematics at University of California, BerkeleyDoctor of Philosophy (Ph.D.), Applied Mathematics, Doctor of Philosophy (Ph.D.), Applied Mathematics at Yale University
Github Skills (11)
compiler-optimization10
pytorch10
machine-learning10
c-language10
cprogramming-language10
testing10
quantization9
cpu9
onnx8
cuda8
python8
Programming languages (6)
Github contributions (5)
pytorch/glow
Jul 2020 - Mar 2022
Compiler for Neural Network hardware accelerators
Role in this project:
Back-end Developer & Test Automation Engineer
Back-end Developer & Test Automation EngineerContributions:17 reviews, 15 commits, 13 PRs in 1 year 7 months
Contributions summary:Huamin primarily contributed to the `glow` repository by fixing and blacklisting tests related to the Neural Network Processing Interface (NNPI). They modified test files, including `NNPIOperatorTest.cpp` and `DeviceManagerTest.cpp`, to address errors and improve testing coverage. Additionally, the user was involved in porting and updating the codebase by incorporating newer versions of NNPI library, and subsequently reverting these changes. Furthermore, they added a flag to the `RecommendationSystemTest` to facilitate dumping model inputs and outputs.
hardware-acceleratorscompilerneural-networkacceleratorshardware
pytorch/pytorch
Jul 2018 - Feb 2022
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:68 reviews, 23 commits, 60 PRs in 3 years 7 months
Contributions summary:Huamin primarily contributed to the AOTInductor compiler within the PyTorch framework, focusing on optimizing code compilation and integration within Meta's production environment. Their work involved addressing compilation issues, refactoring, and enabling new features related to the AOTInductor, including the implementation of quantized linear operators. They also addressed issues related to the integration of custom operators and improved the functionality of the AOTInductor for CPU backends.
pythongpu-accelerationdeep-learninggpunumpy
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.