Zixuan Wei
Software Engineer at 壁仞科技
Minhang District, Shanghai, China
Summary
👤
Senior🎓
Top SchoolZixuan Wei is a software engineer with 11 years' experience specializing in machine learning, deep learning and high-performance computing, currently based in Minhang, Shanghai. He has applied his performance-focused expertise at Intel and in the semiconductor sector, optimizing core ML primitives and MKL-DNN integrations to improve numerical stability and runtime efficiency. At Apache MXNet he contributed backend optimizations—multinomial sampling, RNN gradient fixes, and convolution gradient checks—that tangibly boost model performance in a widely used open-source deep learning framework. He holds an MEng in Control Science and Engineering and combines strong systems-level thinking with practical engineering across Python, Go and C++ stacks. Notably, his background in mechatronics and robotics gives him an uncommon cross-disciplinary perspective on low-level optimization for ML workloads.
11 years of coding experience
2 years of employment as a software developer
Bachelor of Engineering - BE, Mechatronics, Robotics, and Automation Engineering, Bachelor of Engineering - BE, Mechatronics, Robotics, and Automation Engineering at Qinghai UniversityMaster of Engineering - MEng, Control Science and Engineering, TOP 20%, Master of Engineering - MEng, Control Science and Engineering, TOP 20% at East China University of Science and Technology
Chinese
Github Skills (18)
c-language10
python10
mxnet10
machine-learning10
rnn-model10
n10
deep-learning10
performance-optimization10
cprogramming-language10
data-structure9
algorithm9
algorithms9
lstm9
data-structures9
convolution9
Programming languages (9)
Github contributions (5)
apache/mxnet
Jun 2019 - Apr 2020
Lightweight, Portable, Flexible Distributed/Mobile Deep Learning with Dynamic, Mutation-aware Dataflow Dep Scheduler; for Python, R, Julia, Scala, Go, Javascript and more
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:16 commits, 31 PRs, 188 comments in 10 months
Contributions summary:Zixuan primarily contributed to the optimization and enhancement of the MXNet library's core functionalities. They focused on improving the efficiency of multinomial distribution sampling and implemented independent gradient requests checks for convolution operations, impacting the performance of deep learning models. Furthermore, the user addressed issues related to RNN bias gradients and enhanced the MKLDNN (Intel MKL-DNN) integration for RNN operations, demonstrating a focus on performance and numerical stability. They also refactored the code and removed unnecessary cases, thereby improving the code quality and addressing code smells.
pythonschedulerdataflowmutationdata-science
zixuanweeei/mkldnnVerboseParser
Jul 2019 - Dec 2020
Contributions:7 commits, 6 pushes, 1 branch in 1 year 4 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.