Linjian Ma
Staff Research Scientist at Meta
Champaign, Illinois, United States
Summary
🤩
Rockstar🎓
Top SchoolLinjian Ma is a research scientist based in Menlo Park with eight years of experience building and optimizing distributed machine learning systems. At Meta and as an active contributor to PyTorch, he focuses on advanced FSDP features—profiling, prefetching, weight freezing, and parameter execution order—showing deep expertise in large-scale training and GPU acceleration. He combines research rigor with hands-on engineering to bridge algorithmic improvements and production-ready tooling. Notably, his contributions to one of the most widely used deep learning frameworks reflect both practical impact and a nuanced understanding of training performance trade-offs.
8 years of coding experience
5 years of employment as a software developer
Master of Engineering - MEng Computer Science, Master of Engineering - MEng Computer Science at University of California, BerkeleyDoctor of Philosophy - PhD Computer Science, Doctor of Philosophy - PhD Computer Science at University of Illinois Urbana-ChampaignBachelor of Engineering - BE Energy Engineering Chu Kochen Honors College, Bachelor of Engineering - BE Energy Engineering Chu Kochen Honors College at Zhejiang University
Github Skills (9)
pytorch10
machine-learning10
fs10
distributed-training10
deep-learning10
python10
gpu9
neural-network9
autograd8
Programming languages (4)
Github contributions (5)
pytorch/pytorch
Jun 2022 - Jul 2022
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
ML Engineer
ML EngineerContributions:57 reviews, 16 commits, 13 PRs in 1 month
Contributions summary:Linjian primarily contributes to the PyTorch repository, focusing on Fully Sharded Data Parallel (FSDP) functionalities. Their work includes debugging, implementing improvements to FSDP's core features such as profiling and prefetching, fixing weight freezing and parameter execution order, and adding forward prefetching options. The user's contributions demonstrate a deep understanding of distributed training and optimization techniques within the PyTorch framework.
pythongpu-accelerationdeep-learninggpunumpy
LinjianMa/pairwise-perturbation
Jul 2018 - Jun 2020
Pairwise Perturbation: an efficient numerical algorithm for alternating least squares in tensor decompositions
Contributions:1 release, 105 commits, 8 PRs in 1 year 11 months
perturbationpairwisetensorleast-squaresalternating-least-squares
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.