Illia Silin
PMTS Software Development Engneer at AMD
Newport Beach, California, United States
Summary
🤩
Rockstar🎓
Top SchoolIllia Silin is a senior software engineer specializing in high-performance computing and GPU acceleration, currently an SMTS Software Development Engineer at AMD. With over 20 years of experience in C/C++, Fortran, Python and MATLAB, he has repeatedly delivered multi‑order-of-magnitude speedups (up to 2000x) by parallelizing code with OpenMP/MPI and offloading kernels to CUDA/HIP/OpenCL. His background spans academia and industry—from writing six‑dimensional Vlasov solvers during a PhD to optimizing MSC Nastran modules and production software—combining deep numerical methods with practical engineering. At AMD he contributes performance engineering to projects like ROCm/composable_kernel, adding CI benchmarks and tuning GEMM/ResNet kernels for multiple targets. He’s adept at integrating third‑party libraries (Intel MKL, CUDA) and managing HPC workloads with Slurm, PBS Pro and containers, enabling reproducible, high‑throughput experiments. Based in Newport Beach, he brings rare domain expertise that bridges research-grade simulation code and production GPU-accelerated systems.
4 years of coding experience
17 years of employment as a software developer
Doctor of Philosophy (PhD) Physics, Doctor of Philosophy (PhD) Physics at Technische Universität BraunschweigLomonosov Moscow State University
Russian, German, Italian, English
Github Skills (12)
hip10
c-language10
cprogramming-language10
performance-optimization10
scripting9
cicd9
shell9
script9
sh9
cmake8
profiling8
roc7
Programming languages (4)
Github contributions (5)
ROCm/composable_kernel
Apr 2022 - Jan 2023
[DEPRECATED] Moved to ROCm/rocm-libraries repo. NOTE: develop branch is maintained as a read-only mirror
Role in this project:
Performance Engineer
Performance EngineerContributions:517 reviews, 93 commits, 838 PRs in 9 months
Contributions summary:Illia primarily focused on enhancing the performance of the `composable_kernel` repository. Their contributions involve compiling CK for various targets, including fixing typos and missed files, and suppressing warnings. The user added and refined benchmark tests within the CI/CD pipeline, incorporating performance testing and data processing to measure the impact of changes. Significant efforts were made to optimize and assess performance metrics, particularly related to GEMM and Resnet50 kernels.
kerneldeep-learningprogramming-modelmachine-learningtensor-learning
ROCm/AITemplate
Oct 2022 - Jan 2023
AITemplate is a Python framework which renders neural network into high performance CUDA/HIP C++ code. Specialized for FP16 TensorCore (NVIDIA GPU) and MatrixCore (AMD GPU) inference.
Contributions:1 release, 3 reviews, 86 commits in 3 months
amdgpucppcudagpuinference
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.