Giuseppe Rossini
SMTS GPU Architect at AMD
Cambridge, England, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolGiuseppe Rossini is a GPU architect and compiler expert with a decade of experience designing and optimizing ML and HPC compiler backends, currently leading GPU architecture work at AMD in Cambridge. He has deep hands-on expertise in LLVM/MLIR and has contributed performance-critical changes—like improved block merging, redundant-argument elimination and non-temporal masked loads—to the widely used llvm-project. Prior roles across Arm, Huawei, and MathWorks underscore his focus on GPU parallelism and machine-learning compiler pipelines, including practical fixes for Triton backend stream-pipeline issues. He holds a PhD from Télécom Paris and combines rigorous academic training with product-facing engineering, often tackling subtle backend correctness-performance tradeoffs that only surface at scale.
10 years of coding experience
10 years of employment as a software developer
Doctor of Philosophy - PhD Computer Systems Networking and Telecommunications, Doctor of Philosophy - PhD Computer Systems Networking and Telecommunications at Télécom ParisMaster of Science (M.Sc.) Computer Engineering, Master of Science (M.Sc.) Computer Engineering at Università degli Studi di Napoli Federico II
Italian, English, French
Github Skills (13)
compiler-optimization10
triton10
code-generation10
c-language10
amdgpu10
backend10
cprogramming-language10
back-end-development10
mlr10
llvm10
cuda8
vectorization8
hip7
Programming languages (6)
Github contributions (5)
llvm/llvm-project
May 2022 - Oct 2024
The LLVM Project is a collection of modular and reusable compiler and toolchain technologies.
Role in this project:
Back-end Developer
Back-end DeveloperContributions:39 reviews, 18 PRs, 8 pushes in 2 years 5 months
Contributions summary:Giuseppe contributed to the LLVM project by focusing on compiler optimization and code generation. Their work involved modifying the MLIR (Multi-Level Intermediate Representation) code, specifically improving block merging and dropping redundant arguments to improve performance. They also introduced and implemented non-temporal support for masked loads. The user demonstrated expertise in compiler infrastructure and optimization techniques.
compilerstechnologiesclangsubmittoolchain
triton-lang/triton
May 2024 - Mar 2025
Development repository for the Triton language and compiler
Role in this project:Backend Engineer
Backend EngineerContributions:99 reviews, 23 PRs, 179 comments in 9 months
Contributions summary:Giuseppe focused on improving the Triton language and compiler, specifically addressing issues related to AMD GPU stream pipeline optimization and backend functionality. Their commits involved fixing upper bounds in stream pipeline epilogues and updating LLVM versions to incorporate necessary features. The user also made changes to disable block merging in specific situations to prevent compiler issues and enhance overall performance. They demonstrated a strong understanding of compiler internals and backend optimization techniques.
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.