Kurman Karabukaev
Staff Machine Learning Engineer at CoVelocity Technologies
Seattle, Washington, United States
Summary
🤩
RockstarKurman Karabukaev is a Staff Machine Learning Engineer and serial founder based in Seattle with 12 years of experience building distributed, production-grade systems and coaching high-performing teams. He combines deep platform and backend expertise—demonstrated by contributions to PyTorch's TorchElastic for distributed training and logging—with product-level leadership from roles at Meta and Amazon. Kurman has repeatedly launched companies and internal initiatives, blending hands-on engineering with architectural decision-making across startups and enterprises. He excels at making complex ML infrastructure reliable and observable, including work on rendezvous/backends and pluggable logging that improves multi-node debugging. Known for a T-shaped skill set, he moves fluidly between DevOps, backend systems, and machine learning deployments while mentoring teams on delivery practices.
12 years of coding experience
23 years of employment as a software developer
English, Russian, Turkish
Github Skills (15)
pytorch10
distributed-systems10
devops10
python10
machine-learning9
tensor9
c-language8
neural-network8
autograd8
cprogramming-language8
gpu8
kubernetes6
docker6
kubernetes-pods6
dockers6
Programming languages (5)
Github contributions (5)
pytorch/pytorch
Nov 2022 - Aug 2024
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
Back-end & DevOps Engineer
Back-end & DevOps EngineerContributions:74 reviews, 18 PRs, 102 comments in 1 year 9 months
Contributions summary:Kurman contributed to the TorchElastic framework, focusing on enhancements for distributed training and improved logging. They implemented features like customizable log prefixes, facilitating better log analysis in a distributed environment. The user also worked on supporting overprovisioning in the C10D-based rendezvous, enabling more flexible resource management. Further contributions include enabling libuv support for TCPStore rendezvous backends and refactoring the logging mechanism to allow for pluggable log specifications.
pythongpu-accelerationdeep-learninggpunumpy
kurman/pytorch
Oct 2023 - Aug 2024
Contributions:81 pushes, 18 branches in 9 months
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.