Evgeni Krimer
Hardware Architect at Google
Israel
Summary
🤩
Rockstar🎓
Top SchoolEvgeni Krimer is a hardware architect at Google with over 25 years in the industry and a Ph.D. in Electrical and Computer Engineering from UT Austin, combining deep academic rigor with production-focused design. Based in Israel, he brings a rare blend of system-level architecture and hands-on implementation rooted in a long academic pedigree from the Technion through doctoral research. At Google he drives hardware architecture efforts, translating complex computer-architecture research into scalable, deployable systems. He also contributes to open-source ML tooling as an MLOps engineer, notably improving SyncBatchNorm testing and multi-GPU reliability in NVIDIA's widely used apex repository. Colleagues know him for meticulous validation work—refactoring tests and integrating them into CI to boost distributed training robustness and performance. His background suggests a pragmatic thinker who bridges research ideas and engineering realities to make high-performance hardware and software work together.
9 years of coding experience
Doctor of Philosophy (PhD), Electrical and Computer Engineering, Doctor of Philosophy (PhD), Electrical and Computer Engineering at The University of Texas at AustinBachelor of Science (BSc), cum laude, Electrical and Computer Engineering, Bachelor of Science (BSc), cum laude, Electrical and Computer Engineering at Technion-Machon Technologi Le' Israel
Github Skills (9)
cuda10
pytorch10
distributed-training10
testing10
python9
continuous-integration8
ccl7
machine-learning6
apex5
Programming languages (5)
Github contributions (5)
NVIDIA/apex
Feb 2019 - Jun 2019
A PyTorch Extension: Tools for easy mixed precision and distributed training in Pytorch
Role in this project:
MLOps Engineer
MLOps EngineerContributions:9 commits, 2 PRs, 7 pushes in 4 months
Contributions summary:Evgeni primarily contributed to testing and improving the `SyncBatchNorm` functionality within the repository, a key component for distributed training in PyTorch. Their work involved creating and refining unit tests for this feature, ensuring its proper function across multiple GPUs and multi-node environments. This included refactoring code, adding test cases, and integrating these tests into the continuous integration pipeline. Further contributions demonstrate their focus on optimizing the `SyncBatchNorm` implementation for performance.
pytorchraymixed-precisiondeep-learningtemporal-data
ekrimer/training
Sep 2019 - Oct 2019
Reference implementations of training benchmarks
Contributions:1 push, 2 branches in 1 month
implementationsbenchmarkingcaffe2deep-learningmachine-learning
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.