Yu-sheng Li
ML Tooling Software Engineer at Waymo
Taipei, Taiwan
Summary
🤩
Rockstar🎓
Top SchoolYu-sheng Li is an ML tooling software engineer with a decade of experience building production ML and system software, currently improving Waymo’s compute stack after shipping GPU fleet tools at Google Cloud. He has a strong applied ML background from developing recommender systems at Dcard and hands-on computer vision work—contributing to well-known projects like EdgeConnect and PI-REC on image inpainting, translation, and GANs. Comfortable across full-stack, research, and infrastructure domains, he has moved from Android development to CV and GPU systems while optimizing data pipelines and training workflows. He combines rigorous academic training from National Taiwan University with practical research stints at RIKEN and Academia Sinica, and a knack for debugging complex ML codebases. Open to mentoring and coffee chats, he blends product-driven impact with deep implementation skills in both model and system-level engineering.
10 years of coding experience
4 years of employment as a software developer
Master of Science - MS Computer Science, Master of Science - MS Computer Science at National Taiwan University
English, Chinese, Mandarin, Japanese
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (15)
computer-vision10
opencv10
debug10
pytorch10
deep-learning10
cgan10
python10
generative-adversarial-network10
dcgan10
image-processing10
datasets9
data-set9
data-handling9
configuration-management9
edge-detection9
Programming languages (6)
Github contributions (5)
youyuge34/PI-REC
Mar 2019 - Apr 2021
:fire: PI-REC: Progressive Image Reconstruction Network With Edge and Color Domain. :fire: 图像翻译，条件GAN，AI绘画
Role in this project:
Full-stack Developer
Full-stack DeveloperContributions:50 commits, 4 PRs, 47 pushes in 2 years
Contributions summary:Yu-sheng primarily contributed to the interactive image drawing tool of PI-REC, enhancing its functionality and user experience. They added features such as saving the binary edge, color domain, and output images, as well as the ability to lighten the output. Furthermore, the user integrated a refinement model and added a command-line option, demonstrating a focus on both the front-end interface and the underlying image processing logic. The code changes also involved modifications to the main application logic and configuration, reflecting a full-stack development approach.
progressive-imagefireimagecomputer-visioncgan
knazeri/edge-connect
Jan 2019 - Jan 2019
EdgeConnect: Structure Guided Image Inpainting using Edge Prediction, ICCV 2019 https://arxiv.org/abs/1901.00212
Role in this project:ML Engineer
ML EngineerContributions:9 commits, 4 PRs, 12 comments in 5 days
Contributions summary:Yu-sheng primarily focused on debugging and improving the codebase related to image inpainting within the EdgeConnect project. Their commits addressed bugs in the training and testing phases, corrected parameter errors, and optimized data loading. They also made adjustments to the dataset processing pipeline, ensuring correct image sorting and mask handling.
iccv-2019guidedimage-inpaintingiccvarxiv
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.