Shengbao Zheng
Research Scientist at Meta
Mountain View, California, United States
Summary
🤩
Rockstar🎓
Top SchoolShengbao Zheng is a Research Scientist at Meta with a PhD in Computer Science from Duke University and a decade of experience building and optimizing distributed systems. He focuses on performance engineering for large-scale ML infrastructure, contributing to PyTorch’s distributed training tooling—enhancing Execution Trace and Kineto profiling to capture process group, NCCL, and communication metadata for better benchmarking and debugging. His background blends academic rigor from five years as a research assistant with practical industry impact, including intern experience in networking at Big Switch Networks. Based in Mountain View, he pairs deep systems knowledge with a knack for making low-level logging and profiling more efficient and actionable. Colleagues rely on him to translate complex distributed behaviors into measurable signals that drive performance improvements.
10 years of coding experience
5 years of employment as a software developer
Doctor of Philosophy (Ph.D.), Computer Science, Doctor of Philosophy (Ph.D.), Computer Science at Duke UniversityBachelor’s Degree, Computer Science, Bachelor’s Degree, Computer Science at Shanghai Jiao Tong University
Github Skills (19)
pytorch10
performance-monitor10
performance-analytics10
distributed-training10
python10
code-profiling10
performance-analysis10
gpu10
profiling10
ccl10
performance-monitoring10
c-language9
deeplearning-ai9
deep-learning9
cprogramming-language9
Programming languages (3)
Github contributions (5)
pytorch/pytorch
Oct 2023 - Aug 2024
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
Back-end Developer & Performance Engineer
Back-end Developer & Performance EngineerContributions:22 reviews, 30 PRs, 2 branches in 10 months
Contributions summary:Shengbao primarily focused on enhancing the PyTorch distributed training infrastructure, specifically contributing to the Execution Trace (ET) and Kineto profiling tools. Their work involved logging process group configurations, communication operations, and NCCL-related data to improve trace analysis and benchmarking. Key contributions include adding features to record global ranks, root ranks for collective operations, and the NCCL version, demonstrating a focus on performance optimization and debugging of distributed training processes. They also made efforts to refactor and improve the efficiency of the logging mechanisms, contributing to overall improvements in the profiling tools.
pythongpu-accelerationdeep-learninggpunumpy
shengbao-zheng/pytorch
Oct 2023 - Jul 2024
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Contributions:70 pushes, 27 branches in 9 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.