Jianwei Yang
Member Of Technical Staff at xAI
Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolJianwei Yang is a research-driven Member of Technical Staff with 12 years of experience building state-of-the-art multimodal and vision AI systems across industry leaders including Microsoft, Meta, and xAI. He has led development of influential models and architectures—such as Focal Transformer, UniCL, X-Decoder, and Phi-3-V variants—that advanced image/video understanding, grounding, and agentic capabilities for UI navigation and robotics. A prolific ML engineer and open-source contributor, his PyTorch implementations for scene graph generation and Faster R-CNN remain widely used by the community. Jianwei combines deep academic training (PhD-level work across Georgia Tech, Virginia Tech, and CAS) with product-focused research, and is known for inventing practical prompting and parsing techniques (Set-of-Mark, OmniParser) that accelerated multimodal agent development.
12 years of coding experience
6 years of employment as a software developer
Doctor of Philosophy - PhD Artificial Intelligence, Doctor of Philosophy - PhD Artificial Intelligence at Georgia Institute of TechnologyDoctor of Philosophy - PhD Artificial Intelligence, Doctor of Philosophy - PhD Artificial Intelligence at Virginia TechBachelor of Engineering - BE, Bachelor of Engineering - BE at Central South UniversityComputer Vision, Computer Vision at Institute of Automation, Chinese Academy of Sciences
Github Skills (12)
model-building10
gpu-programming10
faster-rcnn10
pytorch10
machine-learning10
computer-vision10
deep-learning10
modeling10
model-driven10
model-driven-development10
back-end-development8
python8
Programming languages (6)
Github contributions (5)
jwyang/graph-rcnn.pytorch
Jul 2018 - Mar 2020
Pytorch code for our ECCV 2018 paper "Graph R-CNN for Scene Graph Generation" and other papers
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:198 commits, 7 PRs, 180 pushes in 1 year 9 months
Contributions summary:Jianwei primarily contributed to the implementation of a scene graph generation model, which involved the creation of a baseline model. The code changes included adding the necessary model, loss function, and data loading components for the project. They worked on defining the model architecture and setting up the training process for the scene graph generation task, utilizing PyTorch for machine learning implementation.
pytorchdeep-learningeccvr-cnneccv-2018
jwyang/faster-rcnn.pytorch
Aug 2017 - May 2020
A faster pytorch implementation of faster r-cnn
Role in this project:Back-end Developer
Back-end DeveloperContributions:312 commits, 49 PRs, 255 pushes in 2 years 10 months
Contributions summary:Jianwei primarily focused on integrating and adapting a PyTorch implementation of Faster R-CNN for GPU processing. Their work involved adding a PyTorch-based non-maximum suppression (NMS) library and modifying the NMS wrapper to support the GPU version. They adapted various RPN and ROI layers to utilize the GPU-accelerated NMS. Additional contributions involved incorporating image trimming and mini-batch loading to the project.
pytorchr-cnnfasterpytorch-implementationfaster-rcnn
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.