Zeyi Wen - Assistant Professor at Hong Kong University of Science and Technology (Guangzhou)

Zeyi Wen

Assistant Professor at Hong Kong University of Science and Technology (Guangzhou)

Guangzhou City, Guangdong Province, China

Join Prog.AI to see contacts

Summary

🤩

Rockstar

🎓

Top School

Zeyi Wen is an Assistant Professor at HKUST (Guangzhou) with 11 years of experience bridging machine learning systems, AutoML, high-performance computing and data mining. He develops production-ready open-source ML systems—most notably ThunderSVM and ThunderGBM—and has contributed deep fixes to ThunderSVM’s GPU/CPU core for improved stability and prediction performance. His work has been recognized with a Best Paper Award at TPDS and an ICPP Best Paper finalist slot, reflecting both theoretical and systems impact. As an Action Editor for JMLR’s MLOSS section, he influences reproducible ML software publication standards. Trained with a PhD from the University of Melbourne and having held roles across top Asia-Pacific institutions, he combines rigorous research with hands-on systems engineering that scales ML algorithms to real-world hardware.

11 years of coding experience

The University of Melbourne

English, Chinese, Chinese, hokkien

Github Skills (14)

algorithm10

cuda10

numerical-optimization10

memory-management10

algorithms10

code-optimization10

regression10

svm10

c-language10

cprogramming-language10

gpu10

optimisation10

classification10

optimization10

Programming languages (2)

C++Python

Github contributions (5)

Xtra-Computing/thundersvm

Dec 2014 - Apr 2022

ThunderSVM: A Fast SVM Library on GPUs and CPUs

Role in this project:

Back-end Developer

Contributions:2 releases, 322 commits, 31 PRs in 7 years 5 months

Contributions summary:Zeyi primarily focused on improving the core functionality of the ThunderSVM library by fixing issues related to Hessian matrix calculations, memory allocation, and prediction functions. The user's contributions addressed bugs in training and prediction functions, and memory allocation within the training process. They also added a function to read kernel values for prediction and refactored the prediction process to separate kernel value memory allocation from the prediction function. These changes contributed significantly to the stability and performance of the SVM library.

cudaregressionlibsvmsvmgpu

Xtra-Computing/thundergbm

Jan 2016 - Jan 2021

ThunderGBM: Fast GBDTs and Random Forests on GPUs

Contributions:1 release, 395 commits, 5 PRs in 5 years

pythongpumachine-learningforestsgpus

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.

Request Free Trial