Sangbum Choi

Grad Student at uw-madison

Madison, Wisconsin, United States
email-iconphone-icongithub-logolinkedin-logotwitter-logostackoverflow-logofacebook-logo
Join Prog.AI to see contacts
email-iconphone-icongithub-logolinkedin-logotwitter-logostackoverflow-logofacebook-logo
Join Prog.AI to see contacts

Summary

🤩
Rockstar
Sangbum Choi is a grad student and experienced software engineer with eight years of hands-on experience in machine learning and systems development, based in Madison, Wisconsin. He contributes to high-impact open-source projects like Hugging Face Transformers, where he has improved DETR-based object detection and segmentation models, fixed auxiliary loss issues, implemented gradient checkpointing, and integrated custom CUDA kernels for performance. Comfortable bridging research and production, he focuses on model reliability, training stability, and code quality while iterating on cutting-edge vision architectures. Known for a methodical, debate-driven approach to problem solving, he values consistency and continuous learning, which shows in both academic work and upstream contributions. Sangbum combines deep technical skill with an eye for practical optimizations that make large models train and run more efficiently.
code8 years of coding experience
github-logo-circle

Github Skills (11)

object-detection10
computer-vision10
pytorch10
machine-learning10
transformer10
nlp10
deep-learning10
dtr10
natural-language-processing10
bert9
cuda9

Programming languages (6)

TypeScriptShellC++JavaScriptJupyter NotebookPython

Github contributions (5)

github-logo-circle
huggingface/transformers

Jun 2023 - Mar 2025

🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
Role in this project:
userML Engineer
Contributions:321 reviews, 26 PRs, 520 comments in 1 year 9 months
Contributions summary:Sangbum primarily contributed to the development and improvement of machine learning models within the Hugging Face Transformers library. Their work included implementing and debugging features related to DETR-based object detection models, such as Deta, Conditional DETR, and Deformable DETR, and the Mask2Former/MaskFormer segmentation models. They focused on fixing auxiliary loss training and gradient checkpointing, as well as improving model training and code quality, and integrating custom CUDA kernels for performance optimization.
pythonbertspeech-recognitionstate-of-the-artflax
SangbumChoi/MobileHumanPose

Oct 2020 - May 2022

Contributions:140 commits, 11 pushes, 3 branches in 1 year 7 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.
Request Free Trial
Sangbum Choi - Grad Student at uw-madison