Daniil Fukalov
SMTS at AMD
Berlin, Germany
Summary
🤩
Rockstar🎓
Top SchoolDaniil Fukalov is a seasoned backend and performance engineer with 11+ years of experience in compiler toolchains and heterogeneous compute tooling, currently working at AMD in Berlin. He has deep expertise in CUDA-to-HIP translation, compiler IR fixes, and build-system automation, contributing significant usability improvements to ROCm/hip’s HIPIFY and optimization-focused patches to LLVM and Triton. Daniil’s work emphasizes memory- and performance-conscious refactors—such as reducing inliner memory usage and correcting APInt handling—and practical tooling enhancements like output filename options and expanded testing. His career spans long tenures at Synopsys and Luxoft/AMD ODC, reflecting sustained impact on complex, production-grade codebases. Trained in computational and applied mathematics at Saint Petersburg State University, he pairs strong theoretical grounding with hands-on systems engineering. An understated strength is his focus on maintainability: many contributions target build reliability and long-term code health rather than flashy features.
11 years of coding experience
9 years of employment as a software developer
Computational and Applied Mathematics, Computational and Applied Mathematics at Saint Petersburg State University
Github Skills (22)
debug10
c-language10
llvm10
build-system10
cmake10
intermediate-code10
hip10
compiler-construction10
compiler-optimization10
performance-tuning10
cuda10
irr10
clangd10
cprogramming-language10
clang10
Programming languages (8)
Github contributions (5)
ROCm/hip
Feb 2016 - Jun 2020
HIP: C++ Heterogeneous-Compute Interface for Portability
Role in this project:
Backend Developer & Automation Engineer
Backend Developer & Automation EngineerContributions:40 commits, 3 PRs, 3 pushes in 4 years 5 months
Contributions summary:Daniil's primary contribution involves the development of the CLANG-based HIPIFY tool for converting CUDA sources to HIP sources. This includes implementing the core logic for translating CUDA code elements into their HIP equivalents. Additionally, the user added and improved testing capabilities and integrated options to the tool, such as the -o option to specify the output filename. These changes indicate a focus on improving the build process, automation, and overall usability of the CUDA to HIP conversion tool.
cudaheterogeneousgpuportabilityhip-runtime
llvm/llvm-project
Oct 2019 - Mar 2025
The LLVM Project is a collection of modular and reusable compiler and toolchain technologies.
Role in this project:Back-end Developer & Performance Engineer
Back-end Developer & Performance EngineerContributions:18 reviews, 26 PRs, 72 pushes in 5 years 5 months
Contributions summary:Daniil primarily contributed to the LLVM project by addressing issues related to the compiler's intermediate representation (IR) and its optimization phases. Their work involved fixing bugs in the `ValueSymbolTable` to correctly handle name size restrictions during inlining. They also refactored the `AlwaysInliner` pass to reduce memory consumption, particularly focusing on removing dead inlined functions to prevent excessive memory usage. Furthermore, the user's contributions extended to improving the efficiency of code generation through the correction of APInt usage, specifically related to offset creation and signed value handling.
compilerstechnologiesclangsubmittoolchain
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.