Hi! My name is Yuichiro Ueno (上野 裕一郎). I am working on computational infrastructure for machine learning workloads.

Work Experience

Full-time Engineer (Apr. 2021 - present)

• Preferred Networks, Inc.
• Machine Learning Platform Engineer

Research Assistant (May 2019 - Mar. 2021)

• AIST-Tokyo Tech RWBC-OIL, AIST
• Scalable Deep Learning

Internship / Part-time Engineer (May 2019 - Jun. 2020)

• Fixstars Corporation
• Development of ClPy, that is an OpenCL extension of NumPy (like CuPy).

Internship (Dec. 2019 - Jan. 2020)

• CyberAgent, Inc.
• Developed a proof-of-concept of Kubernetes CNI using SRv6 (Segment Routing IPv6).
• Blog post: Segment Routingを用いたKubernetes Podネットワーキングの実装

Internship / Part-time Engineer (Aug. 2017 - Mar. 2019)

• Preferred Networks, Inc.
• Implemented an AllReduce collective communication algorithm for InfiniBand / NVIDIA GPU cluster. AllReduce is used for gradient accumulation in data-parallel distributed deep learning.
• Blog post: Technologies behind Distributed Deep Learning: AllReduce

Publications (all peer-reviewed)

• Yuichiro Ueno, Kazuki Osawa, Yohei Tsuji, Akira Naruse, Rio Yokota, Rich Information is Affordable: A Systematic Performance Analysis of Second-order Optimization Using K-FAC, Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, Aug. 2020.

  • DOI

• Kazuki Osawa, Yohei Tsuji, Yuichiro Ueno, Akira Naruse, Chuan-Sheng Foo, Rio Yokota, Scalable and Practical Natural Gradient for Large-Scale Deep Learning, IEEE Transactions on Pattern Analysis and Machine Intelligence, Jun. 2020.

  • DOI

• Yohei Tsuji, Kazuki Osawa, Yuichiro Ueno, Akira Naruse, Rio Yokota, Satoshi Matsuoka, Performance Optimizations and Analysis of Distributed Deep Learning with Approximated Second-Order Optimization Method, Proceedings of the 48th International Conference on Parallel Processing: Workshops, No. 21, Aug. 2019.

  • DOI

• Kazuki Osawa, Yohei Tsuji, Yuichiro Ueno, Akira Naruse, Rio Yokota, Satoshi Matsuoka. Large-Scale Distributed Second-Order Optimization Using Kronecker-Factored Approximate Curvature for Deep Convolutional Neural Networks, IEEE/CVF Conference on Computer Vision and Pattern Recognition, Jun. 2019.

  • DOI, Open Access PDF

• Yuichiro Ueno, Rio Yokota. Exhaustive Study of Hierarchical AllReduce Patterns for Large Messages Between GPUs, 19th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGRID), May. 2019.

  • DOI

Education

Master of Engineering in Computer Science (Apr. 2019 - Mar. 2021)

• Tokyo Institute of Technology
• Advisor: Dr. Rio Yokota
• Theme: Scalable Deep Learning
• GPA: 4.25 / 4.50.

Bachelor of Engineering in Computer Science (Apr. 2017 - Mar. 2019)

• Tokyo Institute of Technology
• Advisor: Dr. Rio Yokota
• Thesis: Exhaustive Study of Hierarchical AllReduce Patterns for Large Messages Between GPUs
• GPA: 3.75 / 4.50.

Associate Degree of Engineering (Apr. 2012 - Mar. 2017)

• National Institute of Technology, Kumamoto College
• Advisor: Dr. Toshinori Nawata
• Thesis: A Study of Distributed Computing Framework Using Fork-Join Model
• Valedictorian of the college.