Oleg Khabinov
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolOleg Khabinov is a seasoned software engineer with a decade of experience enabling AI/ML inference, hardware acceleration, and real-time systems, based in San Francisco. He brings deep hands-on expertise in compiler and backend engineering for neural network accelerators, contributing to high-profile open-source projects like PyTorch's Glow, TensorRT integrations, and AOTInductor. Oleg's work spans CI/DevOps fixes, importer/exporter operator support, CUDA integration, and model packaging—demonstrating both low-level performance tuning and robust production tooling. He is skilled at diagnosing cross-platform build and device issues and adding operator-level support that broadens framework compatibility. A Master’s graduate from Novosibirsk State University, he pairs strong academic foundations with practical impact on widely used ML infrastructure.
10 years of coding experience
Master, Computer Science, Master, Computer Science at Novosibirsk State University (NSU)
Github Skills (20)
pytorch10
caffe10
c-language10
python10
compiler-development10
tensorrt10
glow10
machine-learning10
onnx10
deeplearning-ai10
deep-learning10
gpu10
cuda10
cprogramming-language10
neural-network9
Programming languages (6)
Github contributions (5)
pytorch/glow
Dec 2020 - Jul 2022
Compiler for Neural Network hardware accelerators
Role in this project:
Back-end Developer & DevOps Engineer
Back-end Developer & DevOps EngineerContributions:45 reviews, 68 commits, 82 PRs in 1 year 7 months
Contributions summary:Oleg primarily contributed to the Glow compiler for neural network hardware accelerators, focusing on fixing build issues and improving Caffe2 model loading capabilities. They addressed critical CI build failures by modifying build scripts and enhanced the operator support within the Caffe2 importer, adding functionality for operators like RMSNorm, ReduceBackSum, ArgMin, and Softplus. Their work involved changes to both importer and exporter logic, indicating a focus on ensuring broader compatibility and functionality within the Glow framework.
hardware-acceleratorscompilerneural-networkacceleratorshardware
pytorch/pytorch
Oct 2020 - Jun 2022
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:ML Engineer
ML EngineerContributions:99 reviews, 21 commits, 42 PRs in 1 year 8 months
Contributions summary:Oleg's commits primarily focused on the AOTInductor feature within the PyTorch project, which is a compiler for PyTorch models. They debugged and refined the elimination of small ACC subgraphs, enhanced the AOTInductor's integration with CUDA, and implemented model runners to avoid using torch_extension. Furthermore, the user added tests for multiple CUDA devices and improved validation of the C++ wrapper code generation, demonstrating expertise in optimizing and testing the AOTInductor compilation process.
pythongpu-accelerationdeep-learninggpunumpy
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.