Zhijing Li
Software Engineer at Meta
Sunnyvale, California, United States
Summary
👤
Senior🎓
Top SchoolZhijing Li is a software engineer and PhD student with 8 years of experience bridging machine learning, compiler/hardware tooling, and performance engineering. Based in Sunnyvale, she helped build AITemplate at Meta—an open-source inference framework that delivers multi-fold speedups over PyTorch and TensorRT—and contributes performance improvements to core PyTorch components. Her research background at Cornell and internships (Facebook, Xilinx, Intel) produced domain-specific IRs and HLS languages (EQueue, Calyx, Dahlia) and cycle-accurate simulation tools that enabled significant speed and efficiency gains. Zhijing combines systems-level pragmatism with formal language design, often optimizing memory movement and concurrency for accelerators, and has a track record of turning compiler and hardware research into production-grade performance wins.
8 years of coding experience
5 years of employment as a software developer
Master of Science - MS, Electrical and computer engineering, 3.9, Master of Science - MS, Electrical and computer engineering, 3.9 at Cornell UniversityBachelor's degree, Electrical and Computer Engineering, 3.5, Bachelor's degree, Electrical and Computer Engineering, 3.5 at Shanghai Jiao Tong University
Github Skills (17)
pytorch10
performance-analytics10
performance-monitor10
python10
multithreading10
performance-measurement10
performance-analysis10
performance-tuning10
performance-monitoring10
c-language9
machine-learning9
deeplearning-ai9
deep-learning9
tensor9
cprogramming-language9
Programming languages (9)
Github contributions (5)
pytorch/pytorch
Nov 2022 - Nov 2022
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:1 commit, 6 PRs, 5 comments in 1 day
Contributions summary:Zhijing's commits focus on optimizing and refactoring core components within the PyTorch framework. They made changes to the `layer_norm` implementation, improving its efficiency by symintifying the layer normalization. Additionally, the user implemented multi-threaded readers for the `getRecord` function to enhance performance when handling large tensors, particularly in the context of reading serialized data. Further commits include backing out changes potentially related to performance optimization.
pythongpu-accelerationdeep-learninggpunumpy
tissue3/EyerissSimulator
Feb 2020 - Mar 2020
Contributions:26 commits, 5 pushes, 4 branches in 12 days
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.