Zhuoran Zhao
Staff Research Scientist, PyTorch at Meta
Menlo Park, California, United States
Summary
🤩
Rockstar🎓
Top SchoolZhuoran Zhao is a Staff Research Scientist at Meta focusing on PyTorch GPU inference runtime and compiler work, with nine years of experience advancing large-scale ML serving and inference. He led GPU enablement and broad adoption for Ads ML serving at Meta before moving to core PyTorch runtime optimizations, bringing production-facing compiler experience to open-source contributions. A PhD in Computer Engineering from UT Austin, he combines systems research with hands-on C++ and Python engineering to improve tensor operations and AOTInductor compilation, including performance and compatibility fixes for AMD hardware like MI300. His work on the high-profile pytorch/pytorch repository centers on low-level performance paths (e.g., torch.cat) and cross-vendor GPU support, bridging research insights with deployable runtime improvements. Based in Menlo Park, he blends distributed inference, compiler internals, and practical performance tuning—often surfacing subtle hardware-specific issues that unlock real-world speedups.
8 years of coding experience
11 years of employment as a software developer
Doctor of Philosophy (Ph.D.) Computer Engineering, Doctor of Philosophy (Ph.D.) Computer Engineering at The University of Texas at AustinBachelor's degree Electrical Engineering, Bachelor's degree Electrical Engineering at Zhejiang University
Github Skills (12)
hip10
pytorch10
machine-learning10
tensor10
gpu10
amd10
autograd9
deep-learning9
c-language8
cuda8
cprogramming-language8
python7
Programming languages (3)
Github contributions (5)
pytorch/pytorch
Aug 2023 - Mar 2025
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
ML Engineer
ML EngineerContributions:11 reviews, 20 PRs, 14 pushes in 1 year 7 months
Contributions summary:Zhuoran primarily contributes to the PyTorch codebase, focusing on optimizations and bug fixes related to tensor operations and the AOTInductor compilation backend. Their work includes improving the `torch.cat` implementation, adding fast paths, and addressing issues specific to AMD hardware, including integrating with HIP. They also work on enabling AOTInductor compilation for AMD and fixing related compilation issues, especially for MI300. These changes are primarily focused on performance and compatibility for different hardware backends.
pythongpu-accelerationdeep-learninggpunumpy
SLAM-Lab/NoSSim
May 2017 - Jun 2019
Network-of-Systems Simulator
Contributions:15 commits, 2 PRs, 11 pushes in 2 years 1 month
simulatorsimulationgazebosim
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.