Preferred Networks

Preferred Networks
Name	Preferred Networks
Type	Private
Industry	Artificial intelligence, Robotics, Deep learning
Founded	2014
Founded place	Tokyo, Japan
Founders	Toru Nishikawa, Daisuke Okanohara, Yutaka Matsuo
Headquarters	Tokyo, Japan
Key people	Toru Nishikawa (CEO), Yutaka Matsuo (Director)
Products	Chainer, PFN Platform, edge AI systems
Num employees	700–1000

Preferred Networks

Preferred Networks is a Tokyo-based private company specializing in applied artificial intelligence and deep learning for industrial and research applications. The company focuses on developing software frameworks, edge computing solutions, and robotics integrations, collaborating with technology firms, automotive manufacturers, and academic institutions. It has played a prominent role in the Asia-Pacific AI ecosystem alongside international players such as Google, NVIDIA, and Microsoft.

History

Founded in 2014 in Tokyo by engineers and researchers including Toru Nishikawa, Daisuke Okanohara, and Yutaka Matsuo, the company emerged during a global surge in deep learning investment driven by breakthroughs at institutions like University of Toronto and University of Montreal. Early momentum was propelled by the release of the Chainer framework, which positioned the firm in the same era as projects from Facebook AI Research, OpenAI, and DeepMind. Strategic collaborations with corporations such as Toyota and Fanuc expanded its industrial footprint. Over time, the company attracted talent from universities including The University of Tokyo, Kyoto University, and Osaka University, and established research ties with international labs at Massachusetts Institute of Technology and University of Cambridge.

Technology and Research Areas

Research activities span deep learning, reinforcement learning, robotics, computer vision, and edge computing. The company developed Chainer, an early dynamic computational graph framework comparable to PyTorch and conceptually influenced by work from Stanford University and Carnegie Mellon University. In reinforcement learning, teams explored applications related to autonomous driving and industrial automation, connecting to research traditions exemplified by DeepMind’s Alpha series and Berkeley Artificial Intelligence Research. Computer vision projects intersected with datasets and benchmarks used by groups at Oxford University and ETH Zürich. Edge AI initiatives targeted deployment on hardware platforms from NVIDIA, Intel, and Arm, and addressed real-time inference challenges similar to efforts by Sony and Panasonic.

Products and Services

Product offerings include AI development frameworks, edge inference platforms, and robotic software stacks. Chainer served developers alongside comparable tools from Google’s TensorFlow and Facebook’s Caffe2 before shifting focus toward commercial platforms. The company provided services for automotive clients such as Toyota Motor Corporation and manufacturing partners like Fanuc, offering system integration and performance optimization on accelerators such as NVIDIA Tesla GPUs. Additional services covered custom research collaborations with laboratories at Rikagaku Kenkyusho-affiliated institutions and cloud integration strategies akin to those offered by Amazon Web Services and Microsoft Azure partners.

Partnerships and Collaborations

Strategic partnerships included alliances with global corporations, academic consortia, and government-supported research programs. Notable industrial collaborations involved Toyota, Fanuc, NTT Data, and electronics firms like Panasonic; these collaborations linked applied research to product development in automotive, factory automation, and consumer electronics. Academic partnerships connected the company with research groups at The University of Tokyo, Kyoto University, and Osaka University, mirroring collaborative models used by IBM Research and Siemens. The company participated in collaborative projects funded through national and regional innovation initiatives that included stakeholders such as New Energy and Industrial Technology Development Organization and other Japanese innovation agencies.

Corporate Structure and Funding

As a privately held company, the firm secured funding through venture rounds, corporate investments, and strategic capital partnerships. Major investors and partners included industrial giants like Toyota and venture funds associated with multinational corporations such as Sony’s corporate venture arm. Capital and in-kind support enabled expansion of engineering teams and research labs, while corporate governance reflected practices seen at other private AI entities including DeepMind prior to acquisition and OpenAI in its early phases. Organizational structure combined research units, product engineering teams, and business development divisions interacting with clients in Asia, Europe, and North America.

Notable Projects and Impact

The company engaged in high-profile projects spanning autonomous vehicle prototyping, collaborative robots for manufacturing, and large-scale model training on specialized hardware. Work on autonomous driving involved close engagement with Toyota Research Institute and tested approaches to perception and control analogous to programs at Waymo and Cruise. Robotics projects with Fanuc integrated AI-driven vision and control, echoing developments at KUKA and ABB in industrial automation. Contributions to open-source ecosystems, especially the Chainer framework, influenced researchers and developers at institutions such as RIKEN and AIST and informed practices used at tech firms including LINE Corporation and Rakuten. The company’s emphasis on edge deployment and energy-efficient inference resonated with initiatives from NTT and consumer electronics manufacturers, shaping discussions about commercial AI adoption across sectors.

Category:Artificial intelligence companies Category:Technology companies of Japan