Peter Han
Software Engineer at Iluvatar CoreX 天数智芯
Shanghai, Shanghai, China
Summary
👤
Senior🎓
Top SchoolPeter Han is a seasoned software engineer based in Shanghai with 10+ years specializing in high-performance computing and deep learning on GPUs. He develops AI chip software stacks and gpGPU compilers, driving GEMM and DNN kernel optimizations using CUDA, CUTLASS, cuDNN and LLVM at Iluvatar CoreX. His background spans robotics, 3D point-cloud processing and autonomous driving from roles at Intel and Hozon, giving him a rare blend of low-level GPU performance tuning and applied perception systems. An active contributor to NVIDIA's CUTLASS—where he improved fp16 tests, fixed GEMM/stream bugs and exposed internal structures—he combines production-grade engineering with rigorous test and verification habits. Known for pragmatic problem solving across C/C++ and Python, he brings deep knowledge of parallel reductions, architecture-specific tuning and compiler-driven optimizations to accelerate DL workloads.
10 years of coding experience
15 years of employment as a software developer
Bachelor of Engineering, Software Engineering, Bachelor of Engineering, Software Engineering at Northwestern Polytechnic University (CN)
English, Japanese
Github Skills (8)
cuda10
c-language10
deep-learning10
cprogramming-language10
gpu10
gemfire9
kernel9
parallel-computing8
Programming languages (9)
Github contributions (5)
NVIDIA/cutlass
Mar 2021 - Apr 2022
CUDA Templates for Linear Algebra Subroutines
Role in this project:
Back-end Developer
Back-end DeveloperContributions:2 reviews, 13 commits, 12 PRs in 1 year 1 month
Contributions summary:Peter primarily contributed to the NVIDIA CUTLASS library, a CUDA template library for linear algebra subroutines. Their work included making internal structures public, adding and verifying tests for fp16, and fixing bugs related to stream handling within GEMM and reduction operations. Further contributions included improving the test suite for specific architectures, particularly around the SM60 and parallel split K mode.
cudacpplinear-algebranvidiamatrix-multiplication
Peter9606/neuralnetwork
Jan 2019 - Apr 2020
Contributions:38 commits, 1 push in 1 year 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.