Phil Culliton
Senior ML Engineer at Google DeepMind
Buffalo-Niagara Falls Area United States
Summary
🤩
RockstarPhil Culliton is a Senior ML Engineer with over two decades of experience in ML/AI research and engineering, currently co-owning and driving Gemma.cpp and the Gemma implementation in llama.cpp at Google DeepMind. He combines deep low-level systems expertise—C++, optimized GEMM implementations and CPU-focused inference engineering—with large-scale ML product and developer-relations leadership honed at Kaggle and Google. His background spans game-engine AI and optimization, production ML systems for healthcare and legal discovery, and successful grant-funded research programs, giving him a rare cross-domain perspective on performance, evaluation, and deployment. An active open-source contributor, he has materially improved Gemma.cpp for multi-billion-parameter models (adding 2B/7B support, quantization fixes, and tiled GEMM optimizations), highlighting his focus on practical, high-performance model inference. Based in the Buffalo–Niagara region, he pairs hands-on algorithmic work with program-level leadership across cross-org launches and TPU/inference strategy.
11 years of coding experience
20 years of employment as a software developer
Github Skills (17)
c-language10
operation10
tensorrt10
machine-learning10
inference-engine10
tensorflow10
tensor10
cprogramming-language10
linear-algebra10
quantization9
optimization8
optimisation8
algorithms8
numerical-optimization8
code-optimization8
Programming languages (3)
Github contributions (5)
google/gemma.cpp
Mar 2024 - Mar 2025
lightweight, standalone C++ inference engine for Google's Gemma models.
Role in this project:
ML Engineer
ML EngineerContributions:1 release, 3 reviews, 6 PRs in 1 year
Contributions summary:Phil primarily contributed to the development of the Gemma C++ inference engine. Their work involved resolving critical issues in layer ordering, reshaping, and quantization, improving the model's functionality for 2B models. They also added support for the 7B model and implemented argument parsing. Furthermore, the user worked on matrix multiplication and testing functions, specifically unrolling / tiling GEMM 4x4 functions.
pculliton/gemma.cpp
Apr 2024 - Feb 2025
lightweight, standalone C++ inference engine for Google's Gemma models.
Contributions:70 pushes, 2 branches in 10 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.