Matthew Barrett
Staff Software Engineer at Axelera AI
Cambridge, England, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolMatthew Barrett is a Staff Software Engineer based in Cambridge with six years of experience building and optimizing machine learning compilers for CPUs, GPUs and custom accelerators. He has driven major features at Axelera AI and contributed deeply to Apache TVM—adding target support, operator implementations and optimization passes that enable real-world ML deployment on diverse hardware. Previously at Arm and OctoML he helped bring Ethos accelerators and Qualcomm Hexagon DSPs into TVM and worked on the next-generation Relax IR for dynamic models. Matthew combines low-level compiler engineering with hardware/software co-design, mentoring teammates while improving performance and memory efficiency across stacks. He’s passionate about making AI faster and greener, and his open-source contributions reflect a practical focus on portability and production readiness.
6 years of coding experience
5 years of employment as a software developer
Natural Sciences Physics, Natural Sciences Physics at University of Cambridge
Github Skills (13)
compiler10
relayer10
machine-learning10
compiler-compiler10
deep-learning10
python10
relay10
tensorflow9
gpu9
cuda9
c-language8
tensorrt8
cprogramming-language8
Programming languages (3)
Github contributions (5)
apache/tvm
Oct 2019 - Nov 2022
Open deep learning compiler stack for cpu, gpu and specialized accelerators
Role in this project:
Back-end & ML Engineer
Back-end & ML EngineerContributions:200 reviews, 53 commits, 101 PRs in 3 years 1 month
Contributions summary:Matthew contributed to the TVM project by implementing and fixing several aspects of the compiler stack. Their work included fixing compilation issues for specific LLVM versions, adding support for a Mali Bifrost target, and addressing a bug in the hybrid nms algorithm. Additionally, the user developed support for the TFLite_Detection_PostProcess operator and added a Merge Composite pass for pattern matching and code generation. This work demonstrates a focus on compiler optimization and machine learning model deployment.
metalvulkancompilertensoropencl
mbaret/tvm
Oct 2019 - Mar 2023
Open deep learning compiler stack for cpu, gpu and specialized accelerators
Contributions:2 reviews, 2 PRs, 203 pushes in 3 years 6 months
cpugpu-programminggpu-accelerationtvmdeep-learning
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.