Yuichiro Kikura
Software Engineer
Tokyo, Japan
Summary
🤩
RockstarYuichiro Kikura is a software engineer based in Tokyo with 11 years of experience building high-performance systems and machine learning tooling. Currently at Indeed, he combines web frontend work with ML and HPC expertise, often optimizing low-level kernels and compiler-like integrations for speed. His notable open-source contributions include adding regularization and plotting features to the Chainer deep learning framework and optimizing WebGPU backends and SGEMM kernels for WebDNN. He excels at bridging research-grade ML techniques with production constraints, delivering both algorithmic improvements and pragmatic developer-facing tools. Colleagues would describe him as a pragmatic performance engineer who pays attention to testing, documentation, and visualization to make complex models more understandable and reliable.
11 years of coding experience
Github Skills (20)
unit-testing10
python10
chainer10
machine-learning10
plot10
numpy10
kernel10
deeplearning-ai10
deep-learning10
webgpu10
performance-optimization10
optimisation10
neural-network10
cuda10
optimization10
Programming languages (8)
Github contributions (5)
mil-tokyo/webdnn
Apr 2017 - May 2018
The Fastest DNN Running Framework on Web Browser
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:2 releases, 956 commits, 547 PRs in 1 year 1 month
Contributions summary:Yuichiro's commits indicate significant contributions to the WebDNN project, focusing on enhancements to the WebGPU backend. Their work includes implementing and optimizing kernel code, specifically for operations such as SGEMM, and adding support for various operators, including average pooling, tensor operations, and ReLU. Furthermore, the user demonstrated an understanding of compiler optimization and integration of advanced features such as layer composition to improve the overall performance and functionality of the WebDNN framework.
browserjavascriptoptimizationfastestwebgpu
chainer/chainer
Jan 2016 - Jan 2017
A flexible framework of neural networks for deep learning
Role in this project:ML Engineer
ML EngineerContributions:8 commits, 2 PRs, 4 comments in 11 months
Contributions summary:Yuichiro implemented and tested a LASSO optimizer hook, adding a regularization technique to the Chainer framework. They also added tests for the new optimizer hook and modified existing code to integrate the tests. Furthermore, the user introduced a PlotReport extension, which enables plotting of training statistics and visualizing model behavior during training. Finally, the user made minor changes to documentation and formatting related to the new PlotReport.
cudapythonmxnetcaffe2flexible-framework
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.