Alberto Pérez
Technical Staff at Liquid AI
Greater Santa Cruz de Tenerife Metropolitan Area Spain
Summary
🤩
Rockstar🎓
Top SchoolAlberto Pérez is a Technical Staff GPU software engineer and PhD with nine years of experience optimizing low-level performance, kernel design, and hardware-accelerated AI across Intel and NVIDIA platforms. He has led SYCL backend work for the widely used llama.cpp project, improving matrix-multiplication paths and stabilizing GPU inference on Intel and NVIDIA hardware. At Codeplay he drove SYCL/DPC++ performance migrations and delivered notable algorithm speedups in oneDPL, while also authoring tooling to simplify CUDA-to-SYCL porting. Comfortable at the intersection of research and production, he blends rigorous energy-efficiency analysis from his PhD with hands-on HPC system administration experience. Colocated in the Canary Islands, he thrives on collaborative problem-solving and building practical, high-performance software that scales across diverse accelerator ecosystems.
8 years of coding experience
8 years of employment as a software developer
Doctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at Universidad de La Laguna
Spanish, English, French
Github Skills (17)
c-language10
algebra10
machine-learning10
sycl10
gm10
cuda10
tensor10
gml10
cprogramming-language10
speech-to-text9
inference9
performance-optimization9
openai9
automatic-differentiation8
large-language-models8
Programming languages (2)
Github contributions (5)
ggml-org/llama.cpp
Feb 2024 - Mar 2025
LLM inference in C/C++
Role in this project:
Back-end Engineer
Back-end EngineerContributions:43 reviews, 18 PRs, 8 pushes in 1 year 1 month
Contributions summary:Alberto primarily contributed to the SYCL backend implementation for the llama.cpp project, focusing on optimizing and correcting matrix multiplication operations. Their work involved fixing build issues, re-enabling and improving the mmvq path for the SYCL CUDA backend, and rerouting permuted matrix multiplications through oneMKL. The user also addressed issues in norm functions and updated support conditions for specific operations, improving the overall performance and stability of the SYCL backend.
ggmlllama
HPC-ULL/eml
Jun 2017 - Dec 2020
Contributions:2 releases, 20 commits, 12 pushes in 3 years 6 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.