Adnán Akhundov
Staff Software Engineer at Meta
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolAdnán Akhundov is a Staff Software Engineer in San Francisco with two decades of software experience and a decade-plus focus on ML compilers and GPU kernel performance. At Meta he drives low-level optimizations across PyTorch 2, Triton, and AITemplate, delivering fused kernels and backend improvements that materially speed matrix ops and autotuning. He pairs rigorous research-class ML background (MSc from TUM) with product and engineering leadership from earlier CTO and director roles, enabling him to bridge prototypes to production-grade systems. A prolific open-source contributor, his work in Triton and PyTorch includes deduplication in LLVM IR and cuBLASLt epilogue fusions that improve real-world CUDA workloads. Known as a lifelong learner, he combines systems-level curiosity with practical performance engineering to push the limits of GPU-driven ML.
9 years of coding experience
19 years of employment as a software developer
Master of Business Administration - MBA International Business, Master of Business Administration - MBA International Business at Maastricht School of ManagementMaster of Science - MS Computer Science, Master of Science - MS Computer Science at Technical University of MunichBachelor of Science - BS Applied Mathematics, Bachelor of Science - BS Applied Mathematics at Baku State University
English, Russian, Azerbaijani, German, Turkish, Spanish
Stackoverflow
1reputation
230reached
0answers
1question
Github Skills (19)
pytorch10
c-language10
back-end-development10
python10
llvm10
matrix-multiplication10
machine-learning10
mlr10
triton10
deep-learning10
gpu10
performance-optimization10
compiler-optimization10
cuda10
tensor10
Programming languages (6)
Github contributions (5)
pytorch/pytorch
Jun 2023 - Dec 2024
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
Backend Developer & ML Engineer
Backend Developer & ML EngineerContributions:480 reviews, 84 PRs, 433 pushes in 1 year 6 months
Contributions summary:Adnán primarily contributed to the optimization and improvement of the PyTorch library, specifically focusing on matrix multiplication and related operations. They enabled and refined fused addmm + GELU epilogue fusion using cuBLASLt, improving performance for CUDA-based operations. Further contributions included the addition of a new path in post_grad.py for replacing addmm + ReLU / GELU activation with the corresponding _addmm_activation call. They also addressed performance issues related to user-defined Triton kernels, including their grid, and various related bug fixes.
pythongpu-accelerationdeep-learninggpunumpy
triton-lang/triton
Oct 2023 - Nov 2024
Development repository for the Triton language and compiler
Role in this project:Back-end Developer
Back-end DeveloperContributions:8 reviews, 6 PRs, 34 comments in 1 year 1 month
Contributions summary:Adnán primarily contributed to the Triton compiler, focusing on backend optimizations and features. Their work involved enhancing the compiler's ability to deduplicate computations in LLVM IR, leading to potential performance improvements in generated code. They also expanded MLIR bindings, enabling out-of-tree analysis of the TTIR module. Furthermore, the user addressed issues related to autotuning, ensuring proper passing of arguments to autotuner hooks.
compilerprogramming-languagecode-generationtriton
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.