Mehdi Goli
Principal Software Engineer- AI Performance at Modular
City of Edinburgh, Scotland, United Kingdom
Summary
👤
Senior🎓
Top SchoolMehdi Goli is a Principal Software Engineer and AI Kernel Architect with 11 years of experience optimizing GPU kernels and accelerating large language models across Intel, NVIDIA, AMD, and custom SoCs. He specializes in high-performance kernel design, AI libraries, and deep framework integrations to deliver scalable, efficient inference on CPUs, GPUs and specialized accelerators. At Codeplay he led SYCL adaptations of CUDA libraries (including CUTLASS) and contributed to open standards projects and multi-million-euro EU consortia to democratize heterogeneous acceleration in the cloud. Mehdi combines hands-on kernel development—FlashAttention, GEMM, Softmax and EVT implementations—with systems-level work enabling oneDNN/oneMKL and SYCL support in major frameworks. Based in Edinburgh, he holds a PhD in distributed and parallel programming and has practical open-source impact, including performance-focused contributions to the Eigen mirror that add SYCL support and finer-grained memory handling. He brings a rare blend of research pedigree, cross-vendor portability expertise, and production-grade performance engineering.
11 years of coding experience
13 years of employment as a software developer
Master of Science Computer Science-Analytical and Scientific Databases, Master of Science Computer Science-Analytical and Scientific Databases at The University of EdinburghDoctor of Philosophy (PhD) Distributed and Parallel Programming, Doctor of Philosophy (PhD) Distributed and Parallel Programming at Robert Gordon UniversityM.Eng Software Engineering, M.Eng Software Engineering at Shahid Beheshti University
English, Persian
Stackoverflow
76reputation
913reached
1answer
0questions
Github Skills (10)
sycl10
c-language10
cprogramming-language10
performance-optimization10
linear-algebra10
cpp9
blas8
matrix8
lapack7
eigen6
Programming languages (6)
Github contributions (5)
eigenteam/eigen-git-mirror
Aug 2018 - Nov 2019
THIS MIRROR IS DEPRECATED -- New url: https://gitlab.com/libeigen/eigen
Role in this project:
Back-end Developer & Performance Engineer
Back-end Developer & Performance EngineerContributions:33 commits in 1 year 4 months
Contributions summary:Mehdi's commits primarily focused on converting existing code to use the `EIGEN_STRONG_INLINE` macro and distinguishing between internal and explicit memory allocation/deallocation, which suggests a focus on code optimization and maintainability. The user also introduced a variadic version of an assert function, indicating a possible contribution to improving debugging capabilities. Additionally, the addition of SYCL-related headers and the enabling of per-device specialization of packetsize suggests contributions towards supporting and optimizing Eigen for heterogeneous computing environments.
lapackeigenvalueslinear-algebrasparse-matrixeigen
codeplaysoftware/cutlass-fork
Nov 2023 - Apr 2025
CUDA Templates for Linear Algebra Subroutines
Contributions:336 reviews, 61 PRs, 46 pushes in 1 year 4 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.