Xuan Zhang
Summary
🤩
RockstarXuan Zhang is a research scientist and PhD candidate in computer science at Johns Hopkins University with two years of professional experience focused on ML systems and multimodal language-vision research. Currently at Meta, Xuan applies compiler- and systems-level expertise to optimize deep learning execution—most notably contributing to PyTorch’s Inductor to reduce peak memory usage, improve debugging, and enforce global ordering via torchrec collectives. Prior internships at Microsoft Research and MSR Asia advanced work in language-model pretraining and sign-language translation, complementing ongoing research in hyper-parameter optimization and neural machine translation. Based in Burlingame, California, Xuan blends academic rigor with production-oriented engineering, frequently moving between research questions and low-level implementation details. Outside work, ask them about their cat.
2 years of coding experience
Github Skills (13)
compiler-optimization10
pytorch10
machine-learning10
deep-learning10
python10
autograd9
gpu9
neural-network9
tensor8
memory-management8
debugging7
debug7
mlops5
Programming languages (2)
Github contributions (5)
pytorch/pytorch
May 2024 - Mar 2025
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
ML Engineer
ML EngineerContributions:64 reviews, 27 PRs, 80 pushes in 10 months
Contributions summary:Xuan's contributions center on optimizing the PyTorch Inductor, a compiler for deep learning models. They fixed debugging functionalities within the Inductor framework, ensuring correct buffer usage and visual representation of model graphs. Further work involved reordering scheduler nodes to reduce peak memory usage during model execution and adding signpost events for improved memory pass logging. Additionally, they added and maintained torchrec collectives to enforce global ordering within the Inductor framework.
pythongpu-accelerationdeep-learninggpunumpy
xuanzhang816/pytorch
Jun 2024 - Oct 2024
Contributions:75 pushes, 23 branches in 4 months
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.