Ouadie Farouki
GPU Software Engineer at Symbolica AI
London, England, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolOuadie Farouki is a GPU software engineer with 4 years of focused experience building high-performance AI and GPGPU libraries, currently working at Symbolica AI in London. He has driven SYCL backend development and performance ports for projects like llama.cpp, oneMKL and CUTLASS, improving precision, datatype support and device/event interfaces for LLM inference on Intel and NVIDIA hardware. At Codeplay he combined research and engineering to integrate SYCL across PyTorch, ONNXRuntime and SYCLDNN, demonstrating both production-grade library work and low-level optimization. Trained in High Performance Computing & AI at Mines Paris - PSL and with an engineering degree in scientific computing, he blends strong numerical rigor with practical software delivery. A detail-oriented contributor, he’s known for fixing subtle dequantization and GEMM batching issues that materially improve inference correctness and flexibility.
4 years of coding experience
5 years of employment as a software developer
CNC Mathematics Physics & Engineering Science, CNC Mathematics Physics & Engineering Science at Classes Préparatoires aux Grandes EcolesMaster professionnel High Performance Computing & AI, Master professionnel High Performance Computing & AI at Mines Paris - PSLDiplôme d'ingénieur Modélisation et Informatique Scientifique, Diplôme d'ingénieur Modélisation et Informatique Scientifique at Mohammadia School of Engineers
French, Arabic, English
Github Skills (24)
performance-monitor10
c-language10
performance-analytics10
matrix10
algebra10
machine-learning10
performance-measurement10
llm10
sycl10
mat10
gm10
performance-analysis10
performance-tuning10
tensor10
performance-monitoring10
Programming languages (3)
Github contributions (5)
ggml-org/llama.cpp
Jan 2024 - Oct 2024
LLM inference in C/C++
Role in this project:
Back-end Developer
Back-end DeveloperContributions:26 reviews, 22 PRs, 11 pushes in 8 months
Contributions summary:Ouadie's contributions primarily involve optimizing and enhancing the SYCL backend for the llama.cpp project. They fixed bugs related to data type handling and dequantization, specifically improving precision and ensuring correct calculations for various data types and matrix operations. Furthermore, they enabled support for more data types within the oneMKL gemm_batch function, increasing the flexibility of matrix operations. The user also updated the SYCL device filtering and added a SYCL backend registry, device, and event interfaces.
ggmlllama
OuadiElfarouki/llvm
May 2024 - Nov 2024
Intel staging area for llvm.org contribution. Home for Intel LLVM-based projects.
Contributions:48 pushes, 5 branches in 5 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.