Mingshu Zhou
Software Engineer at IBM
Yangpu District, Shanghai, China
Summary
🤩
RockstarMingshu Zhou is a software engineer with seven years of experience, currently at IBM in Shanghai and pursuing CS research as a PhD student affiliated with thu-pacman. He brings practical expertise in machine learning infrastructure and PyTorch, contributing bug fixes and robustness improvements to fastmoe—a high-performance Mixture-of-Experts implementation—where he resolved gating, type, and communication issues. Comfortable working across research and production boundaries, he focuses on correctness and performance in distributed ML systems. Colleagues would describe him as a detail-oriented engineer who surfaces non-obvious edge cases (like parameter-pop skipping and group communication bugs) that prevent subtle failures at scale.
7 years of coding experience
Github Skills (5)
pytorch10
machine-learning10
python10
distributed-training9
cuda8
Programming languages (5)
Github contributions (5)
laekov/fastmoe
Mar 2022 - Nov 2022
A fast MoE impl for PyTorch
Role in this project:
ML Engineer
ML EngineerContributions:17 reviews, 9 commits, 10 PRs in 7 months
Contributions summary:Mingshu primarily contributed to bug fixes and improvements within the `fastmoe` repository, which focuses on a fast MoE implementation for PyTorch. Their work involved resolving issues related to the GShardGate, type mismatches, and shape inconsistencies. Additionally, the user added a condition to skip parameter popping and fixed a bug related to a proper communication group, showcasing their involvement in optimizing the library's functionality and correctness.
pytorchmoemixture-of-expertsimplpytorch-lightning
thu-pacman/GraphPi
Oct 2019 - Jan 2021
Contributions:123 commits, 4 pushes, 1 branch in 1 year 3 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.