Joe Hattori
Software Engineer at Google
Tokyo, Japan
Summary
👤
Senior🎓
Top SchoolJoe Hattori is a software engineer based in Tokyo with nine years of experience building high-performance systems, currently focused on Android performance at Google. He has deep compiler and ML infrastructure expertise, having worked on model compiler parallelization at Preferred Networks and as a compiler engineer at Meta AI and PFN. His open-source contributions include substantive shape-inference work for the widely-used ONNX standard and practical improvements across Optuna and CuPy, reflecting both low-level systems rigor and attention to developer experience. Joe’s background blends academic research and industry internships—spanning The University of Tokyo, UBC, and Google internships—informing his work on kernel bug detection and virtualization. He’s comfortable bridging research and production, shipping tests, linters, and formatting tools to raise code quality as well as subtle fixes that improve numerical and performance behavior. A detail-oriented engineer, he often focuses on the less-visible but critical areas—shape inference, parallel execution, and tooling—that make ML and systems software reliable at scale.
9 years of coding experience
3 years of employment as a software developer
University of TokyoBachelor's degree, Information Science, Bachelor's degree, Information Science at 東京大学Exchange, Computer Engineering, Exchange, Computer Engineering at University of Toronto
Japanese, English
Github Skills (29)
python10
testing10
machine-learning10
inference10
onnx10
numpy10
hyperparameter-optimization10
cupy10
shapes10
documentation10
basics9
learn-ruby-on-rails9
user-manual9
gpu9
deep-learning9
Programming languages (6)
Github contributions (5)
onnx/onnx
May 2021 - May 2022
Open standard for machine learning interoperability
Role in this project:
ML Engineer
ML EngineerContributions:5 reviews, 9 commits, 11 PRs in 1 year
Contributions summary:Joe primarily contributed to the shape inference capabilities within the ONNX framework, a standard for machine learning interoperability. His work included implementing shape inference for operators like ConstantOfShape and Expand, utilizing symbolic inputs and data propagation techniques. He also fixed a bug in the Resize shape inference and added and updated various tests to ensure shape inference functions correctly. Furthermore, Joe added tools such as flake8 and mypy and clang-format to ensure that the code style is up to standard.
pytorchmxnetdeep-learninginteroperabilitymachine-learning
optuna/optuna
Dec 2021 - Jan 2022
A hyperparameter optimization framework
Role in this project:Back-end Developer / Technical Writer
Back-end Developer / Technical WriterContributions:1 review, 16 commits, 3 PRs in 23 days
Contributions summary:Joe primarily focused on improving the Optuna library's documentation and adding tutorial examples. Their contributions included adding links to the documentation, providing more detailed explanations within existing code, and introducing examples utilizing visualization features like Matplotlib. They also addressed a bug in the TPESampler implementation, demonstrating debugging skills within the hyperparameter optimization framework. Minor formatting changes were included throughout the commits.
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.