Preferred Networks

We are a startup with ~200 members based in Tokyo, focusing on developing technology for Autonomous Driving, Robotics, and Bio-Healthcare. We love open source and are actively developing CuPy and the deep learning framework Chainer. We have one of the world’s largest private GPU clusters, with 1024 NVIDIA Tesla P100 and 512 V100 32Gb GPUs, and we are expanding it further with 1,024 NVIDIA V100 Tensor Core GPUs in July. Using our cluster we achieved the world record at that time of training on ImageNet in 15 minutes. We have a strong emphasis on research, and our technical advisors are Pieter Abbeel (UC Berkeley), Takeo Igarashi (the University of Tokyo), Kenji Fukumizu (the Institute of Statistical Mathematics) and Yarin Gal (University of Oxford). We continuously publish at top conferences like ICRA, ICCV, CVPR, ICLR, ICML, NIPS, NAACL, CHI, UIST, etc.