Kaiyu Yang
Chief Scientist, Verifiable AI Lab at MiroMind.ai
San Francisco Bay Area United States
Summary
🤩
Rockstar🎓
Top SchoolKaiyu Yang is a research-driven AI scientist and engineering leader with 11 years of experience, currently heading the Verifiable AI Lab at MiroMind.ai in the San Francisco Bay Area. He holds a PhD in Computer Science from Princeton and combines deep academic training with industry research stints at Meta and Caltech to advance trustworthy, verifiable machine learning. Kaiyu’s background spans foundational ML systems work—he contributed CUDA-accelerated locally-connected layers to the well-known torch/cunn repository—through to applied verification research, giving him a rare blend of low-level performance engineering and high-level algorithmic rigor. Comfortable moving between code, proofs, and publications, he focuses on making AI both performant and auditable for real-world deployment.
11 years of coding experience
3 years of employment as a software developer
Master of Science - MS, Computer Science, Master of Science - MS, Computer Science at University of MichiganVisiting Student, Computer Science, Visiting Student, Computer Science at Technical University of MunichBachelor of Engineering - BE, Computer Science, Bachelor of Engineering - BE, Computer Science at Tsinghua UniversityDoctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at Princeton University
Chinese, English
Github Skills (9)
neural-network10
cuda10
gpu-programming10
pytorch10
convolutional-neural-networks10
deep-learning10
cprogramming-language9
machine-learning9
c-language9
Programming languages (14)
Github contributions (5)
torch/cunn
Dec 2015 - Apr 2016
Role in this project:
ML Engineer
ML EngineerContributions:6 commits, 4 PRs, 6 comments in 3 months
Contributions summary:Kaiyu primarily contributed to the GPU implementation for locally-connected layers within the `cunn` repository, which suggests a focus on accelerating neural network computations. Their work involved initial CUDA support, including the addition of startup code, addressing compilation errors, and implementing and debugging tests. Significant code changes focused on the `SpatialConvolutionLocal.cu` file, indicating deep involvement in the CUDA-accelerated implementation of spatial convolution operations, with related changes to test cases, and debugging of several related issues.
princeton-vl/attach-juxtapose-parser
Nov 2020 - Nov 2022
Code for the paper "Strongly Incremental Constituency Parsing with Graph Neural Networks"
Contributions:16 commits, 2 PRs, 27 pushes in 1 year 11 months
incrementalconstituencyneural-graphneural-networksparsing
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.