Ilia Taraban
Software Development Engineer at Tenstorrent
Gdańsk, Pomeranian Voivodeship, Poland
Summary
🤩
Rockstar🎓
Top SchoolIlia Taraban is a software development engineer with 12 years of experience specializing in performance engineering and low-level compute for CPUs and GPUs. He has driven optimized implementations and datatype support (notably bfloat16 and u8) in high-profile open-source projects such as oneDNN and DGL, and helped harden LLVM test suites with CPU feature detection and AVX512 tests. His tenure at Intel involved delivering optimized GPU compute stacks (clGPU, clDNN, oneDNN) and compiler stability testing, followed by roles at AMD and Tenstorrent focused on accelerating ML workloads. Ilia combines systems-level rigor with practical engineering, routinely touching both core algorithmic kernels and build/test automation. He holds a Specialist degree in Mathematics and Computer Software from Lomonosov MSU and is based in Gdańsk, Poland. An understated strength is his knack for making advanced numeric types reliable across diverse CPU architectures, avoiding subtle runtime pitfalls.
12 years of coding experience
9 years of employment as a software developer
Lomonosov Moscow State University
Github Skills (28)
pytorch10
c-language10
intrinsics10
python10
autoconf10
avx10
build-system10
cmake10
c1110
standard-library10
c1710
deep-learning10
deep-neural-networks10
performance-optimization10
c-library10
Programming languages (4)
Github contributions (5)
dmlc/dgl
Oct 2022 - Jan 2025
Python package built to ease deep learning on graph, on top of existing DL frameworks.
Role in this project:
ML Engineer
ML EngineerContributions:20 reviews, 11 PRs, 3 pushes in 2 years 3 months
Contributions summary:Ilia contributed to the development of deep learning on graphs by adding and enabling bfloat16 support for CPU operations within the DGL framework. They modified core files related to array operations (SpMM, SDDMM) to accommodate bfloat16 data types. Furthermore, the user enabled bfloat16 conversion functions within the Python API, making it easier for users to utilize the feature. They also contributed to disabling bfloat16 functionality on specific CPUs to avoid unexpected behavior.
pytorchpythondeep-learningmachine-learninggraph-neural-networks
llvm/llvm-test-suite
Dec 2017 - Jan 2019
Role in this project:Back-end Developer & Test Automation Engineer
Back-end Developer & Test Automation EngineerContributions:6 commits in 1 year
Contributions summary:Ilia contributed to the `llvm-test-suite` repository by implementing and enhancing CPU feature detection in the configuration and build processes. Their work involved adding support for CPU architecture detection within the CMake configuration and adding tests using AVX512 intrinsics to evaluate code correctness. The user's contributions focused on modifying build scripts and writing tests to ensure the test suite's compatibility across various CPU architectures and feature sets.
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.