Kunchang Li
Research Intern at 中国科学院深圳先进技术研究院
Beijing, China
Summary
🤩
Rockstar🎓
Top SchoolKunchang Li is a research-focused software engineer with eight years of experience specializing in video understanding and multi-modal foundation models. Currently a research intern at Shanghai AI Laboratory and a PhD candidate at the Shenzhen Institutes of Advanced Technology, he builds efficient and lightweight video models with production-minded experimentation. He has contributed to notable open-source projects such as the UniFormer ICLR2022 implementation, handling dataset/debugging, model loading, and downstream token-labeling tasks across video benchmarks. Past internships at SenseTime and MEGVII reflect hands-on work in vision transformers, knowledge distillation, and recommendation codebases. Based in Beijing, he pairs strong academic performance (PhD studies with a 3.73 GPA) with practical engineering—often modifying configs and run scripts to scale experiments. An interesting thread across his roles is a consistent focus on making cutting-edge video research reproducible and deployable.
8 years of coding experience
博士, 计算机应用技术, GPA 3.73 / 4.0, 博士, 计算机应用技术, GPA 3.73 / 4.0 at University of Chinese Academy of Sciences学士, 计算机软件工程, GPA 3.85/4.0, 学士, 计算机软件工程, GPA 3.85/4.0 at 北京航空航天大学
Chinese, Chinese, Chinese
Github Skills (17)
pytorch10
python10
classification10
image-classification10
machine-learning10
classify10
trainings10
computer-vision10
modeling10
semantic-segmentation7
pose-estimation7
object-detection7
kubernetes-pods5
dockers5
kubernetes5
Programming languages (3)
Github contributions (5)
Sense-X/UniFormer
Jan 2022 - Dec 2022
[ICLR2022] official implementation of UniFormer
Role in this project:
ML Engineer
ML EngineerContributions:34 commits, 4 PRs, 34 pushes in 11 months
Contributions summary:Kunchang appears to be primarily involved in the development and debugging of image classification and video classification models within the UniFormer project. Their contributions include initializing the repository, debugging datasets and model loading, and adding downstream tasks related to token labeling. The user also modifies configuration files and run scripts for various experiments, suggesting they are involved in model training and evaluation across different video datasets and model configurations.
pytorchvision-transformersemantic-segmentationbackbonedeep-learning
OpenGVLab/UniFormerV2
Nov 2022 - Mar 2023
[ICCV2023] UniFormerV2: Spatiotemporal Learning by Arming Image ViTs with Video UniFormer
Contributions:11 commits, 14 pushes, 115 comments in 3 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.