Deqing Sun
Senior Staff Research Scientist And Manager at Harvard University
Cambridge, Massachusetts, United States
Summary
🤩
Rockstar🎓
Top SchoolDeqing Sun is a Senior Staff Research Scientist and manager at Google DeepMind with eight years of experience bridging cutting-edge research and production ML engineering. He holds a PhD in Computer Science from Brown and has a strong research pedigree from roles at Harvard, NVIDIA, and Microsoft Research, where he focused on computer vision and deep learning. Deqing has hands-on experience fixing and adapting influential open-source models—such as contributing practical fixes to the widely cited PWC-Net optical flow repo—to ensure compatibility and reliable model loading across environments. At DeepMind he leads teams translating novel research into robust systems, and his background as both an academic visitor and industry scientist gives him fluency across theory, implementation, and deployment. Colleagues know him for spotting subtle engineering pitfalls (e.g., data-type and deconvolution issues) that commonly break research code when scaled. Based in Cambridge, MA, he combines rigorous scientific thinking with pragmatic engineering to deliver reproducible ML solutions.
8 years of coding experience
4 years of employment as a software developer
PhD Computer Science, PhD Computer Science at Brown UniversityThe Chinese University of Hong Kong (CUHK)
Github Skills (12)
computer-vision10
pytorch10
python10
caffe9
deep-learning8
preloading8
data-loading8
faster-rcnn8
mask-rcnn8
data-processing8
resource-loading8
debug7
Programming languages (2)
Github contributions (5)
NVlabs/PWC-Net
Jun 2018 - Aug 2022
PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volume, CVPR 2018 (Oral)
Role in this project:
ML Engineer
ML EngineerContributions:30 commits, 5 PRs, 32 pushes in 4 years 2 months
Contributions summary:Deqing contributed to the PWC-Net repository by modifying existing code and addressing issues related to model loading and data processing. Their work included converting data types for Python 3 compatibility and adding back a deconvolution layer to resolve a model loading error. Additionally, the user fixed a bug in the image processing script related to sanity checking and incorporated Caffe training protocols for flyingchairs.
pytorchpyramiddeep-learningcnnsoptical-flow
deqings/deqings.github.io
Mar 2019 - Sep 2024
Contributions:17 pushes, 1 branch in 5 years 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.