Masahiro Tanaka

About Masahiro Tanaka

I have been working on distributed computing as a researcher for over ten years. The important projects in my career are the following two:[Development of Middleware for Distributed Deep Learning]- Started a project for training very large-scale neural networks from scratch and developed RaNNC, a middleware that automates the hybrid model/data parallelism- RaNNC is available on GitHub with examples of training billion-scale parameter models- Won first place at PyTorch Annual Hackathon 2021 (officially held by Facebook. 1,947 people participated and 65 projects were submitted)- The research result of RaNNC was published at a top-level conference in parallel/distributed computing (IPDPS 2021)- Have been running RaNNC with hundreds of GPUs to train billion-scale parameter models.[Development of Large-scale NLP applications]- Have been developing large-scale NLP applications including a question-answering system that analyzes over 20 billion Web pages- Developed a new middleware for large-scale orchestration of over 100 resources (NLP tools, NN models, and databases) and ran it on hundreds of compute nodes- The optimization algorithm implemented in the middleware was published at a top-level conference in parallel/distributed computing (IPDPS 2017, CCGrid 2016).

Experience

1. Software Engineer

   Anyscale

   Aug 2025 — Present

Education

• Kyoto University

  Ph.D., Informatics

  2005 — 2009

Skills

• Artificial Intelligence
• Latex
• Computer Science
• Programming
• Linux