Xiaomeng Yang
Research Engineer at Google DeepMind
San Francisco Bay Area United States
Summary
🤩
Rockstar🎓
Top SchoolXiaomeng Yang is a research engineer in the San Francisco Bay Area with 12 years of experience building ML and AI systems at DeepMind, Meta, and Google. She has deep hands-on expertise in reinforcement learning, GPU-accelerated deep learning primitives, and production-grade ML infrastructure, with notable contributions to Caffe2 and PyTorch RL including GPU FCTransposed kernels and optimized batched GEMM operations. Her work spans research and production—improving core performance and robustness in widely used open-source ML frameworks—while also shaping RL training pipelines and data structures. A former PhD student in aeronautics who pivoted to machine learning and computer vision, she blends strong academic grounding with practical system-level problem solving. Colleagues describe her as a pragmatic engineer who surfaces subtle performance bugs and delivers high-impact optimizations that scale across real-world ML workloads.
12 years of coding experience
9 years of employment as a software developer
PhD student Dropped out Aeronautical and Astronautical Science and Technology (Focus on Machine Learning & Computer Vision), PhD student Dropped out Aeronautical and Astronautical Science and Technology (Focus on Machine Learning & Computer Vision) at Tsinghua UniversityBachelor's degree Software Engineering, Bachelor's degree Software Engineering at Beijing Institute of Technology
Github Skills (14)
cuda10
gpu-programming10
pytorch10
machine-learning10
c-language10
deep-learning10
tensorflow10
segment-tree10
cprogramming-language10
python10
ai10
reinforcement-learning10
linear-algebra9
testing8
Programming languages (3)
Github contributions (5)
facebookarchive/caffe2
Dec 2017 - Mar 2018
Caffe2 is a lightweight, modular, and scalable deep learning framework.
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:38 commits, 22 PRs, 6 pushes in 3 months
Contributions summary:Xiaomeng contributed to the implementation of the `FCTransposed` gradient and related tests, including the GPU implementation, which is a core component for deep learning frameworks. The user also worked on improving the performance of `GemmBatchedOp` and other utility functions, likely to optimize the underlying linear algebra operations used in machine learning models. Moreover, the user fixed a bug in a GPU test related to `FCTransposed`, indicating their involvement in testing and debugging GPU-accelerated components.
pytorchscalablecaffe2deep-learningml
pytorch/rl
May 2022 - Jul 2022
A modular, primitive-first, python-first PyTorch library for Reinforcement Learning.
Role in this project:ML Engineer
ML EngineerContributions:2 reviews, 7 commits, 4 PRs in 2 months
Contributions summary:Xiaomeng contributed to the PyTorch RL library by implementing and refining core components related to reinforcement learning. This included syncing SegmentTree implementations from RLMeta to TorchRL, modifying data structures to include support for different data types, and fixing a bug related to segment tree size. They also updated the trainer class, showing contributions to model training pipeline, further demonstrating expertise within the repository's AI/ML context.
pytorchpythonreinforcement-learningprimitivedeep-reinforcement-learning
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.