Leo Fang
Principal System Software Engineer at NVIDIA
New York City Metropolitan Area United States
Summary
🤩
Rockstar🎓
Top SchoolLeo Fang is a Principal System Software Engineer and computational scientist with 11 years of experience specializing in high-performance Python, CUDA, and C/C++ systems. He leads Python CUDA and math library efforts at NVIDIA, driving production GPU software such as cuQuantum and nvmath-python while maintaining core projects like CuPy. Trained as a theoretical physicist (Ph.D. from Duke), he pairs deep domain knowledge in quantum information and optics with practical expertise in code optimization, Monte Carlo, PDE solvers, and numerical integration. An active open-source contributor, Leo has improved foundational projects across the PyData and HPC ecosystem—Numba, mpi4py, DLPack, and conda-forge—often focusing on CUDA-aware interfaces and package compatibility. He bridges research and engineering, having transitioned quantum optics and waveguide QED research into scalable HPC tools and production-grade GPU libraries. Based in the New York City area, he is known for shipping robust, well-tested GPU features that enable both scientific discovery and enterprise-grade performance.
11 years of coding experience
14 years of employment as a software developer
Doctor of Philosophy (Ph.D.), Physics, Doctor of Philosophy (Ph.D.), Physics at Duke UniversityBachelor’s Degree, Physics, Bachelor’s Degree, Physics at National Taiwan University
Stackoverflow
843reputation
52kreached
22answers
4questions
Github Skills (37)
dependency-management10
conda-forge10
github-ci10
restructuredtext10
continuous-integration10
python10
testing10
mpi10
rs10
conda10
microsoft-azure10
cicd10
numpy10
compiler-compiler10
deep-learning10
Programming languages (25)
Github contributions (5)
dmlc/dlpack
Feb 2021 - Dec 2022
common in-memory tensor structure
Role in this project:
Back-end Developer
Back-end DeveloperContributions:37 reviews, 5 commits, 16 PRs in 1 year 10 months
Contributions summary:Leo's contributions primarily involved expanding the DLPack specification. They added support for complex number datatypes, incorporating this feature into the core structure. Furthermore, the user added device types `kDLROCMHost` and `kDLCUDAManaged`, suggesting a focus on enabling support for ROCm and CUDA managed memory. The user also contributed to the documentation by including links to community projects that utilize DLPack. Lastly, they introduced the `kDLBool` type.
tensoroperatordeep-learning
conda-forge/conda-smithy
Apr 2020 - Dec 2022
The tool for managing conda-forge feedstocks.
Role in this project:DevOps Engineer
DevOps EngineerContributions:10 reviews, 26 commits, 5 PRs in 2 years 8 months
Contributions summary:Leo focused on enhancing the continuous integration and continuous deployment (CI/CD) processes for the conda-smithy project. Their contributions included adding support for self-hosted Azure agents, fixing syntax errors, and addressing formatting issues. They also implemented support for the `os_version` configuration within the `conda-forge.yml` file and made various adjustments to the Azure settings and configuration files.
conda-forgecontinuous-integration
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.