Ivan Sidorenko
Senior Software Engineer at Deelvin Solutions Inc.
Moscow, Moscow City, Russia
Summary
🤩
Rockstar🎓
Top SchoolIvan Sidorenko is a senior software engineer based in Moscow with deep expertise in compilers, binary translation, and NPU/GPU toolchains built over a 15+ year engineering career. He has led compiler and NPU projects at Samsung and Huawei and currently develops the TVM compiler at Deelvin, contributing performance-critical QNN optimizations for the popular Apache TVM stack and Hexagon targets. His work spans low-level optimizer design (vectorizers, IPOs, instruction selection) to production post-link binary optimization and hardware-aware cost models, delivering measurable speedups on specialized accelerators. Ivan combines hands-on implementation—new TOPI/QNN operators and AArch64 passes—with team leadership and cross-site coordination, having led a 10-engineer compiler team. Notably, he bridges research and product: publishing practical performance analyses and shipping compiler passes used in silicon-aware deployments. He holds a master’s in applied mathematics and physics from MIPT, reflecting a strong theoretical foundation behind his systems-level engineering.
4 years of coding experience
15 years of employment as a software developer
Магистр, Прикладная математика и физика, Магистр, Прикладная математика и физика at Московский Физико-Технический Институт (Государственный Университет) (МФТИ)
Русский, Английский
Github Skills (7)
compiler-optimization10
hex10
machine-learning10
hexagonal-architecture10
deep-learning9
python8
compiler7
Programming languages (2)
Github contributions (5)
apache/tvm
May 2022 - Jan 2023
Open deep learning compiler stack for cpu, gpu and specialized accelerators
Role in this project:
ML Engineer
ML EngineerContributions:57 reviews, 13 commits, 31 PRs in 8 months
Contributions summary:Ivan implemented and optimized components for quantized neural networks (QNN) within the TVM compiler stack, specifically focusing on the Hexagon architecture. Their work involved enabling and optimizing constant folding for QNN operations, adding new TOPI operators for QNN, and enhancing the performance of existing ones like `nn.pad` and `fixed_point_multiply`. These changes improved performance across various models and provided significant speedups on the target hardware. Furthermore, they introduced a new `qnn.contrib_dense_pack` operation.
metalvulkancompilertensoropencl
ibsidorenko/tvm
Apr 2022 - Sep 2024
Open deep learning compiler stack for cpu, gpu and specialized accelerators
Contributions:99 pushes, 54 branches in 2 years 4 months
cpugpu-programminggpu-accelerationtvmdeep-learning
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.