Shucai Xiao
Principal Member Of Technical Staff at AMD
Austin, Texas, United States
Summary
🤩
Rockstar🎓
Top SchoolShucai Xiao is a Principal Member of Technical Staff with 11 years of hands-on experience building and optimizing GPU software, runtimes, and deep learning inference engines. Based in Austin, he combines low-level C/C++ and Python expertise with parallel and GPU programming across CUDA, ROCm, OpenCL, HIP and graphics APIs to deliver high-performance kernels and drivers. He has driven integration and performance work on MIGraphX and the ONNX Runtime and recently contributed critical AMD backend fixes and MFMA support to the Triton compiler, improving correctness and throughput on AMD GPUs. His background spans production GPU runtime features (debugging, context switch, printf handling) to kernel fusion and transformer inference optimizations, reflecting both systems-level and ML-focused breadth. Known for turning research ideas into production code, he has a PhD-level pedigree from Virginia Tech and a track record of reducing development cycles through tooling and build-system improvements. Colleagues rely on him for complex cross-stack performance problem solving that bridges hardware quirks and compiler/runtime design.
11 years of coding experience
21 years of employment as a software developer
Ph.D Computer Engineering, Ph.D Computer Engineering at Virginia TechB.S. Electronics and Information Technology, B.S. Electronics and Information Technology at Beijing University of Posts and TelecommunicationsM.S. Information and Communication Engineering, M.S. Information and Communication Engineering at Tsinghua University
Github Skills (22)
c-language10
back-end-development10
llvm10
atomics10
gpu-programming10
machine-learning10
onnx10
hardware-acceleration10
deeplearning-ai10
roc10
deep-learning10
atomic10
amdgpu10
backend10
cprogramming-language10
Programming languages (4)
Github contributions (5)
microsoft/onnxruntime
Jun 2020 - Mar 2022
ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator
Role in this project:
ML Engineer
ML EngineerContributions:22 reviews, 9 commits, 12 PRs in 1 year 9 months
Contributions summary:Shucai primarily contributed to the implementation and improvement of the MiGraphX execution provider within the ONNX Runtime. Their work involved adding support for new ONNX operators, including PRelu and batchnorm, as well as refactoring and optimizing existing code. They also addressed build errors, fixed merge conflicts, and conducted code cleanup tasks related to the MiGraphX provider, enhancing its functionality and stability within the ONNX Runtime framework. Further improvements were made to support ROCM including updating to ROCM 4.5.2.
runtimetrainingtensorflowai-frameworkaccelerator
triton-lang/triton
Apr 2023 - Mar 2025
Development repository for the Triton language and compiler
Role in this project:Backend Developer
Backend DeveloperContributions:52 reviews, 37 PRs, 85 comments in 1 year 11 months
Contributions summary:Shucai primarily contributed to the AMD backend of the Triton compiler, focusing on bug fixes and optimization of unit tests. They addressed issues in atomic operations, dot product implementations, and type conversions. Their work involved modifying and improving the code related to TritonGPUToLLVM, specifically for AMD GPUs, including the MFMA (Matrix-Fused Multiply-Accumulate) instructions and the handling of atomic memory operations. The contributions aimed to enhance the correctness and performance of the Triton compiler for AMD hardware.
compilerprogramming-languagecode-generationtriton
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.