Haocong Wang
Software Engineer at 字节跳动
Pudong, Shanghai, China
Summary
🤩
Rockstar🎓
Top SchoolHaocong Wang is a software engineer with six years of experience specializing in GPU performance engineering and high-performance kernel development. He drove operator and kernel optimizations across AMD architectures (RDNA2–RDNA4, CDNA1–CDNA4), notably leading multi-precision matrix multiplication work on MI300 to accelerate AI inference. At AMD he combined back-end development, INT8 support and 3D convolution fixes with a focus on composable kernels that abstract hardware complexity and improve compiler interactions. Now based in Pudong, Shanghai, he continues to apply low-level performance expertise at ByteDance while maintaining contributions to prominent open-source projects like AMD’s MIOpen. His academic background in CFD (Tokyo Tech) and process engineering (Zhejiang University) informs a methodical, physics-aware approach to numerical and performance problems. Colleagues value him for turning deep hardware knowledge into pragmatic, measurable speedups for real-world workloads.
6 years of coding experience
4 years of employment as a software developer
Master of Engineering - MEng, CFD, Master of Engineering - MEng, CFD at Tokyo Institute of TechnologyBachelor of Engineering - BE, Process equipment and control engineering, Bachelor of Engineering - BE, Process equipment and control engineering at 浙江大学
Github Skills (16)
tensorrt10
cuda10
convolution10
hip10
c-language10
tensor10
tensorflow10
cprogramming-language10
performance-optimization10
operation10
ml9
linear-algebra9
machine-learning9
assembly8
assembler8
Programming languages (6)
Github contributions (5)
ROCm/MIOpen
Jan 2022 - Jan 2023
AMD's Machine Intelligence Library
Role in this project:
Back-end Developer & Performance Engineer
Back-end Developer & Performance EngineerContributions:84 reviews, 106 commits, 26 PRs in 11 months
Contributions summary:Haocong primarily contributed to the development and optimization of kernels within the AMD's Machine Intelligence Library. Their work involved implementing tensor reorder features, which included writing optimized code for different tensor layouts and improving overall performance. Furthermore, the user addressed bugs and refactored code for INT8 support, demonstrating involvement in improving the library's functionality and efficiency. The contributions also include adding testing and fixing 3D convolution API for better performance.
amdcluster-computingmachine-learningblasmachine-intelligence
aska-0096/MIOpen
Jan 2022 - May 2022
AMD's Machine Intelligence Library
Contributions:38 pushes, 2 branches in 4 months
amdmachine-learningaimachine-intelligenceintelligence
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.