Tatyana Primak
Deep Learning Manager at Intel Corporation
Hillsboro, Oregon, United States
Summary
🤩
Rockstar🎓
Top SchoolTatyana Primak is a Deep Learning Manager at Intel with nine years of experience applying applied mathematics to high-performance neural network engineering. Based in Hillsboro, Oregon, she blends technical leadership with hands-on performance optimization, notably contributing to the oneDNN open-source library by unifying and vectorizing pooling kernels for AVX2 and AVX512 to boost cross-platform inference efficiency. Her background in applied mathematics informs a pragmatic approach to algorithmic optimization and testing, where she expands test coverage alongside kernel improvements. Tatyana is comfortable operating at the intersection of research-quality numerical methods and production-grade systems, mentoring teams to deliver measurable performance gains. Colleagues rely on her ability to translate low-level architecture features into robust, portable deep learning primitives.
9 years of coding experience
Master's degree, Applied Mathematics, Master's degree, Applied Mathematics at Novosibirsk State Technical University (NSTU)
Github Skills (13)
neural-network10
avx10
c-language10
deep-learning10
edn10
cprogramming-language10
deep-neural-networks10
performance-optimization10
standard-library9
x869
x86-649
c-library9
common-library9
Programming languages (2)
Github contributions (5)
uxlfoundation/oneDNN
Feb 2017 - Jan 2023
oneAPI Deep Neural Network Library (oneDNN)
Role in this project:
Back-end Developer / Performance Engineer
Back-end Developer / Performance EngineerContributions:46 releases, 35 reviews, 100 commits in 5 years 11 months
Contributions summary:Tatyana primarily focused on optimizing the oneAPI Deep Neural Network Library (oneDNN) by implementing and unifying pooling operations for different processor architectures. Their work involved adding support for AVX2 and AVX512 instruction sets, improving code efficiency, and ensuring compatibility across various platforms. The contributions included modifying and extending existing kernel code to leverage advanced vectorization techniques for enhanced performance. Furthermore, the user extended the testing suite to include a broader range of test cases and configurations.
bfloat16sse41avx512openmpopencl
pytorch/pytorch
Jan 2020 - Oct 2020
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Contributions:2 comments in 8 months
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.