Gregory Comer
Bellevue, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolGregory Comer is a software engineer with 10 years of experience specializing in systems-level optimization for ML inference, currently driving PyTorch Edge performance at Meta. He builds high-performance ML compute kernels, compiler/graph optimizations, and runtime improvements—having led the rollout of ExecuTorch/PyTorch 2.0 across Facebook, Instagram, WhatsApp, and Messenger. His background spans productionizing ML/statistical models at Amazon and years of embedded firmware and driver development for precision handhelds, giving him a rare full-stack perspective from microcontrollers to ML frameworks. Gregory is an active open-source contributor focused on integrating performance-critical libraries like XNNPACK into the widely used pytorch/pytorch repo. He frequently collaborates with hardware partners (ARM, Qualcomm, Google) and research teams to co-design small, latency-sensitive models for real-time audio, video, and multi-modal inference. Based in Bellevue, WA, he combines deep systems expertise with a strong emphasis on user experience and adoption.
10 years of coding experience
7 years of employment as a software developer
Bachelor’s Degree, Computer Engineering, Bachelor’s Degree, Computer Engineering at Utah State University
Github Skills (11)
xpack10
pytorch10
machine-learning10
python10
build-automation10
tensor9
compiler8
autograd8
gpu7
deep-learning7
neural-network7
Programming languages (5)
Github contributions (5)
pytorch/pytorch
Dec 2023 - Feb 2025
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
MLOps Engineer
MLOps EngineerContributions:9 reviews, 38 PRs, 16 branches in 1 year 2 months
Contributions summary:Gregory primarily focuses on updating and maintaining the XNNPACK library integration within PyTorch. Their contributions involve updating XNNPACK dependency versions, and modifying build configurations. They also generate and manage operator variants, and adjust the decomposition behavior of ATen ops related to upsampling operations. The user's work demonstrates a strong focus on build processes and the integration of performance-critical libraries within PyTorch.
pythongpu-accelerationdeep-learninggpunumpy
GregoryComer/x86-csv
Sep 2017 - Mar 2021
A machine-readable representation of the Intel x86 Instruction Set Reference.
Contributions:7 commits, 1 PR, 6 pushes in 3 years 6 months
representationreadableintelmsrx86
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.