Ruilong Li
Berkeley, California, United States
Summary
🤩
Rockstar🎓
Top SchoolRuilong Li is a research-focused machine learning engineer with nine years of experience bridging cutting-edge computer vision research and production systems, currently working as a Research Scientist at NVIDIA after PhD work at UC Berkeley and USC. He has a track record of shipping real-time, mobile-first vision features—such as a CNN hair matting algorithm integrated into ByteDance’s FaceU—and delivering performance-focused contributions to high-profile open-source projects like nerfstudio, where he optimized ray-sampling kernels and added CUDA-backed density grids. His background spans academic depth in CV and CG and practical engineering across startups and industry labs, including fast image matching and centimeter-level localization for autonomous systems. Comfortable with both low-level CUDA/PyTorch optimizations and product-driven model design, he combines research rigor with a knack for squeezing latency and quality gains from complex rendering pipelines.
9 years of coding experience
University of California, BerkeleyDoctor of Philosophy - PhD, CV & CG, Doctor of Philosophy - PhD, CV & CG at University of Southern CaliforniaMaster of Science - MS, Computer Science and Technology, Master of Science - MS, Computer Science and Technology at Tsinghua University
English, Chinese
Github Skills (10)
cuda10
raytracing10
pytorch10
computer-vision10
serf10
3d9
3d-reconstruction9
machine-learning9
deep-learning9
3d-graphics8
Programming languages (4)
Github contributions (5)
nerfstudio-project/nerfstudio
Jul 2022 - Dec 2022
A collaboration friendly studio for NeRFs
Role in this project:
ML Engineer
ML EngineerContributions:6 reviews, 20 commits, 26 PRs in 5 months
Contributions summary:Ruilong primarily focused on optimizing the `PDFSampler` implementation within the `nerfstudio` project, specifically targeting performance improvements. Their commits reveal a deep understanding of ray sampling techniques, including the use of `torch.searchsorted` for faster bin calculations. They also addressed issues related to axis arguments in PyTorch functions and updated Instant-NGP and Density Grid with CUDA support to enhance rendering speed and PSNR. Further contributions involved bug fixes and improvements to various samplers within the project.
pytorchphotogrammetrydeep-learningcomputer-vision3d-reconstruction
nerfstudio-project/gsplat
Oct 2023 - Mar 2025
CUDA accelerated rasterization of gaussian splatting
Contributions:16 releases, 154 reviews, 204 PRs in 1 year 5 months
gaussian-splatting
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.