Kai Hu
Pittsburgh, Pennsylvania, United States
Summary
🤩
Rockstar🎓
Top SchoolKai Hu is a research engineer and doctoral student at Carnegie Mellon with eight years of experience building and shipping machine learning research, particularly in video understanding and safety. Currently at Meta and simultaneously pursuing a PhD in ECE, he brings hands-on research experience from internships at Meta, Amazon, ByteDance and Shanghai AI Lab, where he helped maintain the widely used MMAction2 video toolbox. His contributions span model engineering, code hygiene, and benchmarking—adding type hints, fixing lint and merge conflicts, and improving tests to make research code production-ready. Kai has a strong academic foundation in mathematics and electrical and computer engineering, and a track record of top-ranked results in activity recognition competitions, signaling both practical impact and rigorous evaluation. Notably, he balances deep research rigor with pragmatic software craftsmanship, making complex video-ML systems reproducible and robust.
8 years of coding experience
1 year of employment as a software developer
Bachelor's Degree, Mathematics, Bachelor's Degree, Mathematics at Fudan UniversityDoctor of Philosophy, Electrical and Computer Engineering, Doctor of Philosophy, Electrical and Computer Engineering at Carnegie Mellon University
Github Skills (10)
openmmlab10
computer-vision10
pytorch10
deep-learning10
action-recognition10
python10
classify9
classification9
type-checking9
typehinting9
Programming languages (8)
Github contributions (5)
open-mmlab/mmaction2
Jun 2022 - Jan 2023
OpenMMLab's Next Generation Video Understanding Toolbox and Benchmark
Role in this project:
ML Engineer
ML EngineerContributions:97 reviews, 67 commits, 136 PRs in 6 months
Contributions summary:Kai primarily contributed to the codebase by adding type hints to the backbone and head modules of the ResNet3d model. They also addressed a typo in a video backend file. Furthermore, the user fixed conflicts in core files and fixed lint errors by refactoring and improving test files.
avax3dvisual-recognitionbenchmarkvideo
hukkai/rescaling
Jun 2020 - Jun 2022
Official implementation for our paper "Is normalization indispensable for training deep neural networks?"
Contributions:11 commits, 1 PR, 8 pushes in 2 years
pytorchdeep-learningneural-networksnormalizationtraining
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.