Peng Sun is a senior AI software engineering leader with 11 years of experience driving GPU AI software at AMD from low-level compiler and runtime work to high-performance LLM inference and framework integrations. As Sr. Director he leads teams owning ROCm’s GPU AI layer—PyTorch, JAX, TensorFlow, ONNX-Runtime, rocMLIR, Triton backend work and production-grade LLM inference stacks like vLLM and MIGraphX. He combines deep technical roots (HIP/ROCm back-end development and documentation) with client-facing technical ownership for strategic AI accounts, enabling multi-GPU/multi-node performance at scale. His background spans academic research in heterogeneous programming models through to shipping optimized kernels, compilers and operator libraries for GenAI workloads. Based in Austin with a PhD in Computer Science, he is known for bridging research-grade compiler/runtime innovation with pragmatic production engineering. He often surfaces value in overlooked areas such as documentation and API correctness—improvements that materially reduce integration friction for downstream framework adopters.
11 years of coding experience
13 years of employment as a software developer
Doctor of Philosophy (PhD) Computer Science, Doctor of Philosophy (PhD) Computer Science at University of Houston
Bachelor Electronic Information Engineering, Bachelor Electronic Information Engineering at Ocean University of China
Master's degree Computer Information Systems, Master's degree Computer Information Systems at University of Houston-Victoria
HIP: C++ Heterogeneous-Compute Interface for Portability
Role in this project:
Back-end Developer & Documentation Specialist
Contributions:150 commits, 8 PRs, 7 pushes in 2 years 6 months
Contributions summary:Peng primarily focused on code maintenance and documentation improvements within the HIP (Heterogeneous-Compute Interface for Portability) repository. Their contributions centered on fixing the tags in header files and correcting typos, especially in API documentation. Furthermore, the user addressed the error codes.
Dockerfiles for the various software layers defined in the ROCm software platform
Contributions:1 review, 69 commits, 38 PRs in 4 years 10 months
definedradeon-open-computerocmdockerradeon
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.