Zixuan Jiang
San Francisco Bay Area United States
Summary
🤩
Rockstar🎓
Top SchoolZixuan Jiang is a Machine Learning Engineer with eight years of experience building scalable ML infrastructure and parallelization strategies for large models and accelerator farms. Based in the San Francisco Bay Area, he led Shardy and GSPMD at Google to make advanced partitioning the default in JAX, enabling efficient training across thousands of TPUs/GPUs and contributing to Gemini pre-training optimizations. His open-source work includes performance-critical contributions to DREAMPlace, implementing and optimizing DCT/DST transforms with CUDA kernels for VLSI placement toolchains. He combines deep research credentials (PhD-level training at UT Austin) with product-scale engineering across Google, DeepMind, Microsoft, and now Apple Foundation Models. Colleagues rely on him for platform-aware ML solutions that bridge compiler/XLA internals and practical distributed training. A less obvious strength: he repeatedly moves ideas from research prototypes into production-grade partitioners and toolkits, not just papers.
8 years of coding experience
1 year of employment as a software developer
Doctor of Philosophy - PhD, Electrical and Computer Engineering, Doctor of Philosophy - PhD, Electrical and Computer Engineering at The University of Texas at AustinBachelor of Engineering - BE, Electronic Information Engineering, Bachelor of Engineering - BE, Electronic Information Engineering at Zhejiang University
Github Skills (19)
pytorch10
c-language10
cuda10
xla10
cprogramming-language10
optimization10
data-structure9
dft9
algorithm9
algorithms9
machine-learning9
deeplearning-ai9
fft9
deep-learning9
data-structures9
Programming languages (4)
Github contributions (5)
limbo018/DREAMPlace
May 2019 - Nov 2020
Deep learning toolkit-enabled VLSI placement
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:131 commits, 4 PRs, 118 pushes in 1 year 6 months
Contributions summary:Zixuan primarily contributed to the implementation of Discrete Cosine Transform (DCT) and Discrete Sine Transform (DST) functionalities, likely for use in the deep learning toolkit. They introduced new files, corrected syntax errors, and refactored existing code within the DCT module. Their work involved the optimization and implementation of these spectral transforms, including CUDA kernel integration, and unit tests. These contributions are crucial for the toolkit's VLSI placement capabilities.
pytorchgpu-accelerationdeep-learningplacementcomputer-vision
JeremieMelo/dct_cuda
Apr 2019 - Jul 2022
Contributions:39 commits, 2 PRs, 1 push in 3 years 3 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.