Bai-cheng Jeng
Devtech Manager at NVIDIA
Taipei, Taiwan
Summary
🤩
Rockstar🎓
Top SchoolBai-cheng Jeng is a DevTech Manager at NVIDIA with 8 years of experience designing high-performance deep learning and heterogeneous computing solutions for semiconductors. He blends hands-on expertise in CUDA/OpenCL, C++, and HPC with product-focused delivery—having built mobile and automotive deep learning toolkits at MediaTek that dramatically improved throughput, power efficiency, and memory footprint. At NVIDIA he progressed from senior engineer to technical lead and now manages developer technology, driving GPU-optimized toolkits and real-world DL applications for chip customers. An active contributor to CuPy, he optimized asynchronous GPU-to-host transfers by enabling pinned-memory outputs and reducing unnecessary array copies, demonstrating a practical focus on low-level performance gains. He holds a Master’s in Computer Science from National Tsing Hua University and is based in Taipei.
8 years of coding experience
6 years of employment as a software developer
Master's Degree, Computer Science, A+, Master's Degree, Computer Science, A+ at National Tsing Hua University
English, Chinese
Github Skills (9)
asynchronous10
cuda10
gpu10
performance-optimization10
python10
cupy10
async10
numpy10
compiler9
Programming languages (7)
Github contributions (5)
cupy/cupy
Jan 2019 - Jan 2019
NumPy & SciPy for GPU
Role in this project:
Back-end Developer
Back-end DeveloperContributions:5 commits, 6 comments, 5 issues in 6 days
Contributions summary:Bai-cheng focused on optimizing the `get` method within the CuPy library, specifically concerning asynchronous data transfer from the GPU to the host. Their contributions involved refactoring the `get` method to accept an output array, allowing for the use of pinned memory for faster data transfers. They also removed the use of `ascontiguousarray` and `asfortranarray` in favor of directly checking the contiguity flags of the output array, and updated the documentation to reflect the changes. These improvements aimed to improve the performance of asynchronous data transfers and reduce overhead.
cudapythoncusolvergpunumpy
jeng1220/cuGemmProf
Dec 2019 - Feb 2021
Contributions:101 commits, 74 pushes, 9 branches in 1 year 1 month
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.