Yuqing Tang
Principal Research Scientist at IDEA
Kirkland, Washington, United States
Summary
🤩
RockstarYuqing Tang is a Principal Research Scientist and roboticist with nine years of industry experience building large-scale autonomy and AI systems, currently leading science and engineering teams at IDEA on lower airspace systems. He combines deep LLM and deep learning expertise with practical robotics and multi-agent planning, spanning distributed sensor networks, CNS+X, large-scale trajectory planning and airspace simulation. Previously he led consumer robotics at XPeng, contributed to Alexa AI and Facebook AI on multilingual and multimodal LLM applications, and worked on foundational deep learning at Microsoft CNTK. An active open-source contributor, his work on projects like CNTK and fairseq improved documentation, multilingual training and memory efficiency—practical fixes that enabled broader adoption. He has published 50+ papers across AI, ML, robotics and symbolic–probabilistic hybrids, and serves on program committees for top conferences including NeurIPS, ICLR, AAAI and AAMAS. His long-term aim is to fuse human-knowledge representations with data-driven learning to endow AI with richer cognitive capabilities.
9 years of coding experience
Chinese, English, Chinese
Github Skills (8)
pytorch10
machine-learning10
artificial-intelligence10
cntk10
deep-learning10
python10
documentation9
back-end-development9
Programming languages (5)
Github contributions (5)
microsoft/CNTK
May 2017 - Oct 2018
Microsoft Cognitive Toolkit (CNTK), an open source deep-learning toolkit
Role in this project:
Back-end Developer
Back-end DeveloperContributions:167 commits, 194 pushes, 130 branches in 1 year 4 months
Contributions summary:Yuqing's commits primarily involve adding simple descriptions to Python packages and modules within the CNTK (Microsoft Cognitive Toolkit) repository. These descriptions aim to clarify the functionality of various learners, operators, and core components like NDArrayView and Value. The user has also fixed grammar and typesetting issues within the module summary docstrings, enhancing the clarity and readability of the documentation. Additionally, the user added examples demonstrating how to create user minibatch sources.
pytorchpythondeep-learningc-plus-plusmachine-learning
facebookresearch/fairseq
Feb 2020 - Jan 2021
Facebook AI Research Sequence-to-Sequence Toolkit written in Python.
Role in this project:Back-end Developer
Back-end DeveloperContributions:81 commits, 5 comments, 4 issues in 11 months
Contributions summary:Yuqing primarily contributed to the `fairseq` project, a sequence-to-sequence toolkit written in Python, focusing on multilingual training and model optimization. Their work included implementing features for multilingual training with multiple bitext and monolingual datasets, enhancing the data loading and sampling mechanisms. The user also fixed bugs related to fp16, and improved the memory efficiency of the system. They introduced support for manifold sharding data and adjusted the training process.
pytorchnlpsequencepythontransformer-architecture
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.