Natalia Gimelshein
Software Engineer at Meta
San Jose, California, United States
Summary
🤩
Rockstar🎓
Top SchoolNatalia Gimelshein is a software engineer in San Jose with a decade of experience specializing in high-performance deep learning and GPU-accelerated systems. Her background spans research and production roles at Meta, OpenAI, and NVIDIA, where she shipped optimizations for core libraries like cuDNN and contributed to PyTorch and the Triton compiler. She focuses on correctness and performance—fixing subtle synchronization and type-promotion bugs, improving numerical tests, and speeding JIT launches and autotuning for kernels. Earlier work in computational science delivered multi‑x GPU speedups and automated data-analysis pipelines for complex physics simulations, reflecting a strong applied-math foundation. That mix of research rigor and production impact makes her adept at turning low-level numerical challenges into reliable, scalable ML infrastructure.
10 years of coding experience
23 years of employment as a software developer
MS Aerospace Aeronautical and Astronautical Engineering, MS Aerospace Aeronautical and Astronautical Engineering at Penn State UniversityMS Environmental Science, MS Environmental Science at Novosibirsk State University (NSU)
Github Skills (18)
pytorch10
operation10
python10
testing10
tensorrt10
triton10
tensorflow10
gpu10
performance-optimization10
compiler-optimization10
cuda10
performance-tuning10
tensor10
c-language9
machine-learning9
Programming languages (9)
Github contributions (5)
pytorch/pytorch
Sep 2019 - Jan 2023
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
Back-end Developer
Back-end DeveloperContributions:4245 reviews, 807 commits, 609 PRs in 3 years 4 months
Contributions summary:Natalia contributed to the PyTorch library, primarily focusing on performance and correctness improvements. Their work involved fixing issues in comparison operations by addressing potential synchronization problems and optimizing the type promotion behavior. The user also contributed to various functional and numerical tests by adding new sample inputs, enabling edge cases, and improving the precision of existing tests. They further worked on enhancements related to the handling of numeric and data types during memory operations.
pythongpu-accelerationdeep-learninggpunumpy
triton-lang/triton
May 2022 - Nov 2022
Development repository for the Triton language and compiler
Role in this project:ML Engineer & Performance Engineer
ML Engineer & Performance EngineerContributions:13 reviews, 6 commits, 13 PRs in 5 months
Contributions summary:Natalia primarily focused on optimizing the performance and functionality of the Triton language and compiler. They improved the launch speed of JIT functions, fixed bool conversions, adjusted heuristics for a critical kernel, fixed broadcasting for the `where` operation, and improved the autotuning cache. These changes were made within the Triton frontend and test environments.
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.