Shunta Saito
Researcher at Preferred Networks, Inc.
Chiyoda, Japan
Summary
🤩
Rockstar🎓
Top SchoolShunta Saito is a Ph.D. researcher and seasoned ML engineer with 14 years of experience, currently working at Preferred Networks in Tokyo. He combines deep academic training from Keio and UC Berkeley with practical system-level expertise gained at Facebook and multiple startups, focusing on computer vision and GPU-accelerated deep learning. An active open-source contributor and Kaggle Master, he has made notable backend and model-level contributions to high-profile projects such as Chainer/ChainerCV, CuPy, and TensorComprehensions, including semantic segmentation models and spatial pyramid pooling implementations. His work spans from writing core CUDA-backed primitives to improving serialization and autotuning for ML compilers, showing strength in both low-level performance engineering and applied research. Colleagues rely on him to bridge prototype research and production-ready tooling, and he often uncovers subtle bugs in tensor handling and evaluation metrics that improve real-world model robustness.
14 years of coding experience
4 years of employment as a software developer
Computer Vision, Computer Vision at University of California, BerkeleyDoctor of Engineering, Computer Science, 2nd, Doctor of Engineering, Computer Science, 2nd at 慶応義塾大学 / Keio University
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (24)
c-language10
convolutional-neural-networks10
python10
model-driven10
chainer10
testing10
semantic-segmentation10
machine-learning10
model-building10
deep-learning10
cupy10
neural-network10
computer-vision10
cprogramming-language10
modeling10
Programming languages (7)
Github contributions (5)
chainer/chainer
Jul 2015 - Jul 2019
A flexible framework of neural networks for deep learning
Role in this project:
Back-end Developer
Back-end DeveloperContributions:11 releases, 972 commits, 319 PRs in 4 years
Contributions summary:Shunta made several commits to fix test cases related to varied margins in a contrastive loss function, and a bug related to the handling of negative margins was resolved. They also merged updates from the upstream branch to resolve conflicts in the chainer functions and documentation. The user addressed issues involving data type and shape mismatches and introduced a new upsampling function.
cudapythonmxnetcaffe2flexible-framework
chainer/chainercv
May 2017 - Sep 2019
ChainerCV: a Library for Deep Learning in Computer Vision
Role in this project:Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:180 commits, 23 PRs, 70 pushes in 2 years 4 months
Contributions summary:Shunta primarily contributed to the development and integration of machine learning models within the ChainerCV library, specifically for computer vision tasks. Their work involved adding new models, such as SegNet and SegNetBasic, along with associated datasets like CamVid, and refining existing modules. The user also focused on implementing semantic segmentation evaluation metrics and training scripts, demonstrating a focus on model development, dataset handling, and practical application within the computer vision domain.
deep-learningpytorchcomputer-visionvision
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.