Jiqun Tu
Senior Compute Performance Developer Technology Engineer at NVIDIA
New York, New York, United States
Summary
🤩
Rockstar🎓
Top SchoolJiqun Tu is a Senior Compute Performance Developer Technology Engineer at NVIDIA with a decade of experience applying high-performance C++ and CUDA to real-world problems. He transitioned from a PhD in lattice QCD at Columbia, where he developed intensive parallel and GPU code, into production performance engineering focused on accelerating data-processing stacks. At NVIDIA he contributes to flagship open-source projects like rapidsai/cudf, adding a generic PTX function interface for binary ops and refactoring kernels to exploit device shared memory for measurable speedups. His work bridges deep systems knowledge and developer ergonomics—creating Python-facing interfaces for low-level PTX kernels and adding benchmarks to prove gains. Based in New York, he combines rigorous academic training with practical engineering that surfaces non-obvious performance wins across GPU data frameworks.
10 years of coding experience
5 years of employment as a software developer
Bachelor’s Degree, Physics, Bachelor’s Degree, Physics at Fudan UniversityDoctor of Philosophy (Ph.D.), Physics: Lattice QCD, Doctor of Philosophy (Ph.D.), Physics: Lattice QCD at Columbia University in the City of New YorkPhysics, Physics at University of California, Los Angeles
English, Chinese
Github Skills (19)
c-language10
dataframes10
cudf10
dataframe10
kernel10
performance-optimization10
kernel-mode10
compile10
ptx10
cuda10
numba10
cprogramming-language10
jit10
benchmark9
python9
Programming languages (4)
Github contributions (5)
rapidsai/cudf
Jun 2019 - Nov 2020
cuDF - GPU DataFrame Library
Role in this project:
Back-end Developer & Performance Engineer
Back-end Developer & Performance EngineerContributions:1 review, 202 commits, 18 PRs in 1 year 5 months
Contributions summary:Jiqun contributed significantly to the cudf library by implementing and optimizing a new generic PTX (Parallel Thread Execution) function interface for binary operations. Their work included adding a new interface that allows users to define custom CUDA kernels through PTX code. This development involved refactoring existing code, updating the PTX parser, and creating a python interface for the new functionality. Furthermore, the user enhanced performance by adding benchmarks to the library, for example, the gather benchmark, while refactoring gather kernel by using device side shared memory.
cudadataframe-librarydata-analysiscppcudf
hummingtree/cps-public
Feb 2017 - May 2019
Contributions:11 pushes, 1 branch in 2 years 2 months
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.