Yuri Putivsky
Retired at Terimber
San Jose, California, United States
Summary
🤩
Rockstar🎓
Top SchoolYuri Putivsky is a seasoned software engineer and former CEO with over 30 years of experience and a decade focused on modern software and ML systems in the San Jose area. He built and led high-performance cross-platform back-end projects at Terimber and architected real-time, large-scale media and ML infrastructure during an 11-year tenure at Facebook, contributing to video, image, speech, and similarity search pipelines. His open-source work includes backend and ML engineering on high-profile projects like PyTorch/Glow—adding gradient support and operator expansions for neural network accelerators—and performance/resource monitoring in Facebook’s Bistro scheduler. A physicist by training (PhD, MSU) who began his career in laser and optics research, Yuri brings rigorous analytic thinking to systems engineering and a knack for squeezing performance from both hardware and software. Even in retirement he remains engaged with low-level optimization and tooling that bridges ML training and production inference.
10 years of coding experience
37 years of employment as a software developer
Lomonosov Moscow State University
Github Skills (19)
pytorch10
c-language10
operation10
image-classification10
tensorrt10
glow10
machine-learning10
onnx10
subprocess10
gradient10
system-monitoring10
deep-learning10
tensorflow10
performance-optimization10
neural-network10
Programming languages (3)
Github contributions (5)
pytorch/glow
Apr 2019 - Jan 2020
Compiler for Neural Network hardware accelerators
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:45 commits, 67 PRs, 8 pushes in 9 months
Contributions summary:Yuri primarily contributed to the `pytorch/glow` repository by implementing and improving the `ImageClassifier` class and related scripts. They focused on testing and optimizing the image classification functionality, as evidenced by changes to the test scripts and configurations. Furthermore, the user added gradient calculations for various nodes, demonstrating a focus on training capabilities, and expanded the support for operators, which are specific to Glow like, `ConstantOfShape`, `BatchedReduceAddNode` and `GatherNode` . This suggests a strong involvement in back-end development and machine learning engineering, particularly concerning the optimization of neural networks.
compilerhardwareneural-network
facebookarchive/bistro
Oct 2015 - Nov 2015
Bistro is a flexible distributed scheduler, a high-performance framework supporting multiple paradigms while retaining ease of configuration, management, and monitoring.
Role in this project:Backend Developer
Backend DeveloperContributions:7 commits in 25 days
Contributions summary:Yuri implemented a system for collecting and reporting subprocess resource usage statistics, including RAM, CPU, and GPU utilization. This involved integrating the Sigar library for system information gathering and extending the `KillableSubprocess` class. The contributions demonstrate a focus on performance monitoring and resource management within the Bistro scheduler framework. The user also refactored parts of the server to improve tests and added command line execution capabilities with timeout functionality.
monitoring
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.