Kevin Chih-yao
Member Of Technical Staff at Microsoft AI
Seattle, Washington, United States
Summary
👤
Senior🎓
Top SchoolChih-Yao Ma is a research scientist and Ph.D. candidate in Electrical and Computer Engineering at Georgia Tech with a decade of experience in computer vision, deep learning, and multimodal video understanding. His work spans image saliency, real-time RGB-D multi-object tracking, human action recognition, video captioning, and vision-and-language navigation, blending reinforcement learning and NLP with visual models. Proficient in Python, C/C++, Lua, and major deep learning frameworks (PyTorch, TensorFlow, MXNet, Caffe), he brings both academic rigor and practical engineering—evidenced by contributions like enhancing video-overlay prediction tooling for activity-recognition models. Based in Atlanta, he now works as a Staff Research Scientist on GenAI at Meta, bridging cutting-edge research and production ML systems. Notably, his background combines long-term research focus with hands-on implementation details that improve reproducibility and demoability of video understanding models.
10 years of coding experience
6 years of employment as a software developer
Master's degree Electrical and Computer Engineering, Master's degree Electrical and Computer Engineering at National Chiao Tung UniversityDoctor of Philosophy (Ph.D.) School of Electrical and Computer Engineering, Doctor of Philosophy (Ph.D.) School of Electrical and Computer Engineering at Georgia Institute of Technology
Chinese, English
Github Skills (10)
computer-vision10
opencv10
machine-learning10
python10
video-understanding9
mask-rcnn9
faster-rcnn9
torch-net8
torch-hub8
torchscript8
Programming languages (7)
Github contributions (5)
chihyaoma/Activity-Recognition-with-CNN-and-RNN
Apr 2016 - Dec 2018
Temporal Segments LSTM and Temporal-Inception for Activity Recognition
Role in this project:
ML Engineer
ML EngineerContributions:217 commits, 174 pushes, 3 branches in 2 years 8 months
Contributions summary:Kevin primarily worked on the `TextOverlay/TextOverlay.py` file, adding functionalities to print predictions on videos. They implemented features to overlay text displaying probabilities, class labels, and ground truth information onto video frames. The user also extended this functionality to generate predictions for multiple videos using a list of predictions stored in text files and incorporated predictions from a temporal CNN.
pytorchlstm-neural-networksdeep-learninglstmrecognition
chihyaoma/selfmonitoring-agent
Jan 2019 - Sep 2021
PyTorch code for ICLR 2019 paper: Self-Monitoring Navigation Agent via Auxiliary Progress Estimation
Contributions:13 commits, 1 PR, 11 pushes in 2 years 8 months
pytorchiclrauxiliaryagenticlr2019
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.