Xunnan Xu
Software Engineer at Meta
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolXunnan Xu is a software engineer with 10 years of experience building large-scale distributed systems and ML infrastructure, now a Member of Technical Staff at a stealth AI startup in San Francisco. Previously at Meta he focused on distributed training resilience, communication optimization (ncclx zero SM comm overlapping), and elastic checkpointing at production scale. He contributes to PyTorch’s core, improving memory and shared-memory support for distributed workloads, reflecting deep low-level systems and ML-runtime expertise. Xunnan combines academic rigor from an MS in Computer Science (USC) with practical experience in cluster scheduling and ad-tech, and favors quiet, high-impact engineering over showy work.
10 years of coding experience
8 years of employment as a software developer
Master of Science (MS) Computer Science, Master of Science (MS) Computer Science at University of Southern CaliforniaBachelor of Science - BS Industrial Engineering, Bachelor of Science - BS Industrial Engineering at Shanghai Jiao Tong University
English, Chinese
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (11)
pytorch10
c-language10
distributed-training10
tensor10
deep-learning10
cprogramming-language10
python9
gpu9
machine-learning9
neural-network8
autograd8
Programming languages (10)
Github contributions (5)
pytorch/pytorch
Dec 2021 - Feb 2025
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
Back-end Developer
Back-end DeveloperContributions:40 reviews, 21 PRs, 13 pushes in 3 years 3 months
Contributions summary:Xunnan primarily contributes to the PyTorch library's core functionalities by implementing and modifying low-level storage and memory management features. Their work includes adding support for shared memory within the ATen library, enabling memory sharing across processes, and integrating new functionalities like the `share_memory_()` method to `TensorBase`. Additionally, the user addresses critical issues by backing out problematic changes and correcting errors in existing code, such as those related to the handling of sharding and placement within the distributed training framework. These contributions focus on enhancing the library's capabilities in distributed training and memory efficiency.
pythongpu-accelerationdeep-learninggpunumpy
xunnanxu/playframework
Mar 2017 - Mar 2023
Play Framework
Contributions:5 pushes, 2 branches in 6 years
javaplay-frameworkframework
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.