Richard Zou
Staff Software Engineer at Meta
New York, New York, United States
Summary
🤩
Rockstar🎓
Top School
Top expert inArtificial Intelligence and Machine Learning Technologies
Richard Zou is a Staff Software Engineer and long-time PyTorch maintainer based in New York with 11 years of experience building high-performance ML infrastructure and tooling. At Meta he co-led functorch, enabling advanced autodiff (jacobians, hessians, per-sample gradients) and making previously costly numerical experiments practical inside PyTorch. His open-source work spans custom C++ operators, batching and vmap rules, CI/builder automation, and front-end documentation—demonstrating full-stack fluency across core framework internals and developer experience. Richard combines deep systems-level expertise with a pragmatic focus on test automation and reproducible builds, evidenced by contributions to pytorch/pytorch, functorch, and the pytorch builder. Trained in computer science at Harvard, he brings both academic rigor and production-hardened engineering to ML platform challenges.
11 years of coding experience
1 year of employment as a software developer
Bachelor of Arts (B.A.) Computer Science, Bachelor of Arts (B.A.) Computer Science at Harvard UniversityColumbia UniversityHigh School Diploma, High School Diploma at Hunter College High School
English
Github Skills (31)
pytorch10
c-language10
continuous-integration10
pytest10
operation10
python10
tensorrt10
machine-learning10
bash10
cicd10
manage10
tensorflow10
front-end-development10
autograd10
html10
Programming languages (11)
Github contributions (5)
pytorch/functorch
Apr 2021 - Jan 2023
functorch is JAX-like composable function transforms for PyTorch.
Role in this project:
Backend Developer
Backend DeveloperContributions:4 releases, 417 reviews, 915 commits in 1 year 8 months
Contributions summary:Richard contributed to the `functorch/functorch` repository, a library for composable function transforms in PyTorch. Their commits primarily involved modifying C++ source files, specifically focusing on the implementation of batching rules for core PyTorch operations such as the backward pass for `_log_softmax`, `mse_loss`, and `index_put`. The changes included refactoring and adding specialized rules to improve performance of vmap operations and address issues with operator composability within the library.
pytorchgradientshessianscomposablejax
pytorch/pytorch
Sep 2017 - Jan 2023
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:ML Engineer
ML EngineerContributions:3 releases, 6016 reviews, 2918 commits in 5 years 4 months
Contributions summary:Richard primarily contributed to the development of custom operators within the PyTorch framework, specifically focusing on enhancing the framework's capabilities with new functionalities and optimizing their performance. The user's contributions included implementing and testing operations related to tensor manipulation, functionalization and autograd formulas, which are essential for improving the capabilities of machine learning libraries. The user made significant additions, and they also demonstrated an understanding of various components within the PyTorch framework, as evidenced by the registration and design of these custom ops.
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.