Adam Osewski
SMTS Software Development Eng. at AMD
Pomeranian Voivodeship, Poland
Summary
🤩
Rockstar🎓
Top SchoolAdam Osewski is a senior software engineer specializing in high-performance parallel algorithms for multicore and manycore architectures, with eight years of experience focused on ML inference and image/content processing. He leads AMD's Composable Kernel efforts in Poland, contributing hands-on to GPU-optimized deep learning primitives and mentoring a local engineering team. Adam has a strong open-source pedigree—implementing ONNX operators and expanding nGraph/PaddlePaddle support for RNN/GRU, bfloat16, and numerous numerics wrappers—demonstrating deep integration between compiler/runtime and hardware-aware kernels. His background spans Intel and research labs, combining low-level C++/CUDA-style optimization with Python tooling, CI, and testing discipline. Colleagues rely on him for bridging algorithmic design and platform-specific engineering to squeeze both correctness and performance out of modern accelerators.
8 years of coding experience
8 years of employment as a software developer
Master of Science (M.Sc.), Computer Science, Master of Science (M.Sc.), Computer Science at Warsaw University of Technology
Polish, English, German
Github Skills (18)
unit-testing10
gru10
c-language10
python10
graph10
rnn-model10
onnx10
n10
deep-learning10
paddlepaddle10
cprogramming-language10
efficientnet10
machine-learning9
neural-network9
computer-vision9
Programming languages (5)
Github contributions (5)
NervanaSystems/ngraph
Mar 2018 - May 2020
nGraph - open source C++ library, compiler and runtime for Deep Learning
Role in this project:
Back-end Developer
Back-end DeveloperContributions:91 commits, 136 PRs, 534 pushes in 2 years 2 months
Contributions summary:Adam primarily contributed to the Python wrappers for nGraph operations within the `nervanasystems/ngraph` repository. Their work involved implementing Python wrappers for a variety of nGraph operations like Cos, Cosh, Reduce, Relu, Sign, Sin, Sinh, Tan, Subtract, Select, Tanh, Sum, ReplaceSlice, Reverse and more. The commits included updates to docstrings, type annotations, and the addition of unit tests to validate the implemented functionalities. Furthermore, the user refactored and improved the signatures and functionalities of existing operations like `avg_pool`.
inference-enginecppc-librarydeep-learningtvm
PaddlePaddle/Paddle
Aug 2020 - Oct 2021
PArallel Distributed Deep LEarning: Machine Learning Framework from Industrial Practice （『飞桨』核心框架，深度学习&机器学习高性能单机、分布式训练和跨平台部署）
Role in this project:Back-end Developer & MLOps Engineer
Back-end Developer & MLOps EngineerContributions:139 reviews, 16 commits, 23 PRs in 1 year 1 month
Contributions summary:Adam primarily contributed to optimizing the PaddlePaddle framework for deep learning, with a focus on performance and efficiency. Their work involved updating dependencies and optimizing kernel subroutines. They also made changes to integrate and utilize BF16 (bfloat16) data type support, adding it to protobuf messages and implementing it in operators, demonstrating an understanding of hardware-level optimization. Further commits include modifications to the fusion passes and checkpoint operator attributes.
pytorchpythonparalleldeep-learningpaddlepaddle
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.