Hiroyuki Yamazaki
Software Engineer, Engineering Manager at Preferred Networks, Inc.
Tokyo, Japan
Summary
🤩
Rockstar🎓
Top SchoolHiroyuki Yamazaki is a software engineer and engineering manager at Preferred Networks with 12 years of experience building compilers and workload optimizations for in-house deep learning accelerators and maintaining production-grade open source ML tooling. He regularly works with PyTorch and ONNX and has contributed backend improvements and refactors to prominent projects like Chainer and CuPy, plus PyTorch Lightning examples and pruning integration for Optuna. As a project lead and former GSoC mentor, he blends hands-on compiler engineering with team leadership, delivering both low-level performance work and higher-level model workflows. Multilingual in English, Japanese, and Swedish, he brings a linguist’s precision to API design and documentation—an unusual asset in systems-focused ML engineering.
12 years of coding experience
9 years of employment as a software developer
Master’s Degree, Computer Science and Engineering Japanese Profile, Master’s Degree, Computer Science and Engineering Japanese Profile at KTH, Royal Institute of TechnologyMaster’s Degree, Computer Science, Master’s Degree, Computer Science at Keio University
English, Japanese, Swedish
Github Skills (18)
c-language10
python10
chainer10
machine-learning10
pytorch-lightning10
numpy10
hyperparameter-optimization10
cprogramming-language10
testing9
pytorch9
backpropagation9
linear-algebra9
deep-learning9
kernel9
cuda9
Programming languages (9)
Github contributions (5)
chainer/chainer
Mar 2016 - Jan 2020
A flexible framework of neural networks for deep learning
Role in this project:
Back-end Developer
Back-end DeveloperContributions:8 releases, 3099 commits, 788 PRs in 3 years 10 months
Contributions summary:Hiroyuki's contributions primarily revolved around refactoring and improving the ChainerX code base. Their work involved checking and correcting the calling of runtime functions, creating unit tests for core functions, and ensuring the functionality and stability of various routines in the system. Their changes include both code structure and test case improvements, covering areas like linear algebra routines and the handling of gradients. They were also instrumental in updating dependencies like gsl-lite.
cudapythonmxnetcaffe2flexible-framework
optuna/optuna
Oct 2019 - Dec 2022
A hyperparameter optimization framework
Role in this project:ML Engineer
ML EngineerContributions:17 releases, 1165 reviews, 1104 commits in 3 years 2 months
Contributions summary:Hiroyuki primarily focused on integrating and demonstrating the usage of PyTorch Lightning within the Optuna framework. They implemented and updated example code utilizing PyTorch Lightning for hyperparameter optimization of neural networks and added the necessary dependencies to support it. The user also introduced a pruning callback specific to PyTorch Lightning, further integrating it into the Optuna workflow.
pythonoptimization-frameworkparallelhyperparameteroptimization
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.