Toshihiro KONDA
Japan
Summary
👤
SeniorToshihiro Konda is a senior compiler engineer and performance analyst with eight years focused on high-performance compilers and DNN codegen, currently leading a small team at ArchiTek building an aIPE compiler (LLVM/clang/lld) and bespoke RISC-V backends. He has a track record of rapidly implementing complex function units and ISA improvements solo—sometimes delivering primitives in days—and has repeatedly optimized VLIW, long-word codegen, and auto-parallelization for AI chips. Earlier roles span low-level compiler backends (PEZY, DeepInsights), TensorFlow XLA work, and hands-on performance tuning at Fujitsu, giving him deep expertise in ISA design, super-wide SIMD, and applied performance analysis. He describes himself more as a performance analyst than a general programmer, which explains his strength in logical root-cause analysis that ties microarchitecture to real-world performance. Notably, he invented the UXSIMD algorithm and has contributed pragmatic innovations like novel rerolling and parallel pipeline techniques that appear across specialized AI compilers. Based in Chiyoda, Japan, he combines research-grade systems experience with rapid, production-focused implementation.
8 years of coding experience
Github Skills (13)
opencl7
javascript6
gpu6
deep-learning5
machine-learning5
vulkan5
tvm5
metal5
spirv5
raspberry-pi4
compiler4
tensor3
llvm2
Programming languages (2)
Github contributions (5)
LEA0317/LLVM-VideoCore4
Jun 2019 - May 2022
LLVM toy compiler for VideoCore4 (Raspberry Pi VPU)
Contributions:144 commits, 152 pushes, 9 comments in 3 years
compilerllvmraspberry-pi
LEA0317/llvm9-test
Nov 2019 - May 2022
test optimization for X86
Contributions:7 PRs, 10 pushes, 1 branch in 2 years 6 months
x86optimization
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.