Ziheng Jiang
Machine Learning Engineer at University of Washington
Seattle, Washington, United States
Summary
🤩
RockstarZiheng Jiang is a Machine Learning Engineer based in Seattle with 10 years of experience building and optimizing ML systems and compiler backends. He contributes to high-profile open-source projects like MXNet and TVM, where his work spans improving imperative APIs, thread-safe profiling, datatype support, and compiler optimizations for CPU/GPU accelerators. Ziheng specializes in backend and performance engineering, frequently touching low-level C++/CUDA code to extend data type support and streamline inference graphs. He has a strong systems-oriented research interest in machine learning infrastructure and a track record of practical refactors that improve robustness and cross-platform compatibility (including Darwin dynamic library handling). Colleagues would note his knack for making subtle low-level changes that unlock broader performance and usability gains across ML toolchains.
10 years of coding experience
Github Skills (28)
performance-monitor10
c-language10
back-end-development10
mxnet10
machine-learning10
deeplearning-ai10
compiler-compiler10
performance-analysis10
deep-learning10
cuda10
compiler10
nvm10
cprogramming-language10
cudnn10
data-structure9
Programming languages (9)
Github contributions (5)
apache/tvm
Apr 2017 - Sep 2022
Open deep learning compiler stack for cpu, gpu and specialized accelerators
Role in this project:
Back-end Developer & Performance Engineer
Back-end Developer & Performance EngineerContributions:2 releases, 77 reviews, 122 commits in 5 years 6 months
Contributions summary:Ziheng's commits focus on optimizing and extending the core compiler stack for deep learning. Their contributions include implementing support for partition loops with thread axes, improving the handling of bitwise operations, and extending the data type support in the compilation flow. The user has also worked on the implementation of the tensor compute op and the simplification of the graph for inference phase. This demonstrates a deep understanding of the system, and a focus on enhancing the core functionalities.
metalvulkancompilertensoropencl
dmlc/nnvm
May 2017 - Jan 2018
Role in this project:ML Engineer
ML EngineerContributions:14 commits, 20 PRs, 4 pushes in 8 months
Contributions summary:Ziheng focused on modifying the nnvm library, which is a neural network graph compiler. Their contributions included refactoring code (renaming types, and making data structures generic), adding comments, and fixing lint issues. Furthermore, the user worked on enabling loading of dynamic libraries on Darwin systems and optimizing the precompute and prune functions. These changes suggest an effort to improve the library's functionality and performance.
cudametalcomputation-graphtvmdeep-learning
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.