Jerry Zhang
Co Founder at Startup
Black Diamond, Washington, United States
Summary
🤩
Rockstar🎓
Top School
Top expert inArtificial Intelligence and Machine Learning Technologies
Jerry Zhang is a PhD-trained machine learning engineer and entrepreneur with 11 years of experience building and optimizing production-ready ML systems, currently serving as Co-Founder and an Engineering Manager at Facebook. He specializes in model quantization and deployment for large language and vision models, contributing significant TorchAO and PyTorch-level improvements (including int4/int8/float8 support, FX graph-mode quantization, and GEMM optimizations) to widely used open-source projects. His background spans research at IBM and Microsoft Research to product-focused engineering at Facebook, giving him a rare blend of deep research rigor and large-scale production experience. Based in Black Diamond, WA, he repeatedly bridges low-level backend changes and developer-facing tooling—often authoring tutorials and utilities that make advanced quantization techniques accessible to practitioners.
11 years of coding experience
4 years of employment as a software developer
Doctor of Philosophy (PhD) Computer Science, Doctor of Philosophy (PhD) Computer Science at University of Illinois Urbana-ChampaignBachelor's degree Computer Science, Bachelor's degree Computer Science at Tsinghua University
Github Skills (34)
pytorch10
quants10
c-language10
caffe10
learn-ruby-on-rails10
python10
user-manual10
machine-learning10
inference10
basics10
llm10
deep-learning10
model-optimization10
cuda10
quantization10
Programming languages (9)
Github contributions (5)
pytorch/pytorch
Apr 2018 - Jan 2023
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
Back-end Developer
Back-end DeveloperContributions:1 release, 3527 reviews, 5178 commits in 4 years 10 months
Contributions summary:Jerry contributed to the PyTorch/PyTorch repository by adding and modifying utility functions related to quantized neural networks, specifically focusing on adding a function to retrieve example inputs for submodules to facilitate quantization. Their work involved modifications to the `torch/ao/quantization/utils.py` file and the creation and testing of a new function. Further, the user addressed problems with existing test cases by skipping tests using hypothesis.
pythongpu-accelerationdeep-learninggpunumpy
pytorch/torchtune
Mar 2024 - Oct 2024
PyTorch native post-training library
Role in this project:ML Engineer
ML EngineerContributions:37 reviews, 17 PRs, 29 pushes in 7 months
Contributions summary:Jerry focused on enhancing the `torchtune` library by adding support for post-training quantization, a technique to optimize the performance of Transformer-based LLMs. Their contributions involved integrating `torchao`, a library specializing in quantization, and modifying core recipe files to load and save quantized models. They also implemented functionality to enable torch.compile for further generation performance speedups. Furthermore, they refactored model saving and added logging for bandwidth achieved during model generation.
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.