Ramin Azarmehr
GPU Software Engineer at Apple
Cupertino, California, United States
Summary
🤩
Rockstar🎓
Top SchoolRamin Azarmehr is a GPU software engineer based in Cupertino with four years of focused experience building high-performance graphics and compute systems. He combines a strong academic foundation (MSc, University of Ottawa) with hands-on expertise in 2D/3D graphics, computer vision, SIMD algorithms and GPU kernel optimization. At Apple he advances GPU software, and his open-source contributions to PyTorch’s MPS backend show practical impact on ML performance and correctness—adding kernels, fixing memory/ dtype bugs, and optimizing linear algebra. Prior roles at EA and BlackBerry QNX reflect a blend of real-time graphics, systems software and embedded vision experience across production game and automotive-grade platforms. Colleagues rely on his problem-solving, prioritization and clear communication to move complex cross-team projects forward. An underrated strength is his track record of translating research-grade algorithms into robust, production-ready GPU implementations.
4 years of coding experience
17 years of employment as a software developer
Master’s Degree, Master’s Degree at University of Ottawa
Persian, English
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (13)
pytorch10
c-language10
cprogramming-language10
back-end-development10
mps10
metal-performance-shaders10
linear-algebra9
performance-optimization9
python7
tensor7
deep-learning7
gpu6
autograd5
Programming languages (2)
Github contributions (5)
pytorch/pytorch
May 2022 - Jan 2023
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:134 reviews, 37 commits, 55 PRs in 7 months
Contributions summary:Ramin's contributions center around enhancing the MPS (Metal Performance Shaders) backend for the PyTorch framework. They implemented and fixed various issues related to MPS support for specific operations, including linear algebra, padding, and random number generation. Their work involved adding new functionalities such as floor division and addressing data type issues to ensure correct behavior, along with fixing bugs related to memory management and backward pass calculations in loss functions. The user significantly improved performance by optimizing the linear algebra kernels and providing custom kernel support.
pythongpu-accelerationdeep-learninggpunumpy
razarmehr/benchmark
Aug 2022 - Mar 2023
TorchBench is a collection of open source benchmarks used to evaluate PyTorch performance.
Contributions:2 reviews, 11 pushes in 7 months
pytorchbenchmarkingevaluatetoolboxtorch
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.