Xinjiang Wang
Principal Researcher at Microsoft
Hong Kong, China
Summary
🤩
Rockstar🎓
Top SchoolXinjiang Wang is a Principal Researcher based in Hong Kong with 11 years of experience bridging deep learning research and production-grade computer vision systems. He has led both academic and industrial work—from theoretical analyses of neural network regularization to high-impact CVPR papers and practical defect-inspection deployments—before driving LLM and RLHF efforts that helped SenseChat rank top in SUPERCLUE. Xinjiang is an active open-source contributor to the prominent OpenMMLab ecosystem (mmdetection, mmcv), improving core assigners, FSAF detection heads, and robustness fixes that affect many downstream users. He combines hands-on backend engineering with research leadership, having built semi-supervised and scale-equalizing detection methods and overseen large-model pretraining and continual learning at scale. An uncommon strength is his dual fluency in foundational theory and pragmatic system hardening, enabling both novel algorithms and stable production tooling.
11 years of coding experience
8 years of employment as a software developer
博士, Mechanical Engineering, A-, 博士, Mechanical Engineering, A- at 香港科技大学学士, 热能与动力, 90.4 (/100), 学士, 热能与动力, 90.4 (/100) at 华中科技大学
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (10)
openmmlab10
object-detection10
computer-vision10
pytorch10
python10
image-processing9
algorithms8
data-structures8
algorithm8
data-structure8
Programming languages (1)
Github contributions (5)
open-mmlab/mmdetection
Apr 2020 - Jan 2022
OpenMMLab Detection Toolbox and Benchmark
Role in this project:
Back-end Developer
Back-end DeveloperContributions:121 reviews, 29 commits, 37 PRs in 1 year 9 months
Contributions summary:Xinjiang's primary contribution involves modifying and improving the assignment process within the object detection framework. They focused on ensuring consistency and correctness in label assignment, specifically fixing background label assignments across multiple assigners. Additionally, the user implemented features, configurations and addressed bugs related to the FSAF (Feature Selective Anchor Free) head, a key component for object detection, demonstrating a focus on core framework enhancements. Further contributions involve fixing bugs, adding configurations and enhancing the robustness of the system.
retinanetbenchmarkfast-rcnnopenmmlabsemisupervised-learning
open-mmlab/mmcv
May 2020 - Apr 2021
OpenMMLab Computer Vision Foundation
Role in this project:Back-end Developer
Back-end DeveloperContributions:34 reviews, 21 commits, 20 PRs in 10 months
Contributions summary:Xinjiang contributed to the core functionality of the mmcv library. Their work included adding features such as a sleep function during epoch transitions to prevent deadlocks. They also added gradient norm logging to the optimizer hook and fixed related bugs. Furthermore, the user implemented a pillow backend for image loading and added support for normal dict checkpoints.
visiondeep-learningcomputer-visionfoundationopenmmlab
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.