Gleb Larochkin
MTS Software Development Engineer at AMD
Cambridge, England, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolGleb Larochkin is an MTS Software Development Engineer specializing in low-level C/C++ and GPGPU software with eight years of experience and a Master's in Computer Engineering. He designs and optimizes high-performance convolution kernels across AMD GPU generations (MI100–MI300, RDNA3–RDNA4), improving real-world ML workloads like Stable Diffusion and FSR by multiple-fold. Comfortable in Assembly, HIP/OpenCL/CUDA-like languages and tooling (CMake, PyTorch, Docker, perf counters), he owns the full optimization lifecycle from micro-benchmarking to integration and accuracy testing. An active contributor to AMD’s open-source MIOpen project, he implemented Winograd solver kernels and architecture-specific fixes that materially boosted library performance. He also has a track record of porting and integrating inference modules for autonomous systems and building debug tooling for GPU assembly development. Based in Cambridge, he combines deep systems-level expertise with practical ML acceleration experience, often finding gains through non-obvious kernel and memory-layout optimizations.
7 years of coding experience
7 years of employment as a software developer
Master's degree Computer Science Engineering, Master's degree Computer Science Engineering at ITMO University
Russian, English
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (8)
c-language10
hpc10
cprogramming-language10
performance-optimization10
amd10
deep-learning8
machine-learning8
cuda5
Programming languages (9)
Github contributions (5)
ROCm/MIOpen
Dec 2020 - Jan 2023
AMD's Machine Intelligence Library
Role in this project:
Back-end Developer & Performance Engineer
Back-end Developer & Performance EngineerContributions:72 reviews, 83 commits, 41 PRs in 2 years 1 month
Contributions summary:Gleb made significant contributions to AMD's Machine Intelligence Library (MIOPEN), focusing on optimizing convolution operations, specifically Winograd solvers for enhanced performance. They updated and implemented new kernels for the Winograd algorithm, including support for new architectures like gfx10. The user also addressed performance bottlenecks, implemented features for big tensor support, and fixed code related to reported compute units.
amdcluster-computingmachine-learningblasmachine-intelligence
Slimakanzer/dotfiles
Nov 2019 - Feb 2023
Dot files for Ubuntu
Contributions:2 PRs, 19 pushes, 3 branches in 3 years 2 months
dotlinuxdot-filesdotfilesubuntu
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.