DEEPX, an on-device AI semiconductor company, is expanding its collaboration with the global AI development ecosystem to target the physical AI market. The strategy is to build an environment that encompasses everything from development to proof of concept (PoC) and industrial application by connecting AI models, developer platforms, and hardware with its NPU.

DeepX announced on the 14th that it is accelerating the establishment of an open physical AI ecosystem by expanding its connection with global AI models and developer platforms.

Physical AI refers to technology in which AI directly performs recognition, judgment, and control on real-world devices such as robots, smart factories, intelligent cameras, and industrial equipment. DeepX plans to support developers in expanding AI developed in familiar environments to actual industrial sites by connecting ultra-low-power NPUs with a diverse ecosystem of AI models.

DeepX is first strengthening its connection with Ultralytics' YOLO ecosystem, which is widely used in the field of vision AI. The company entered into a strategic partnership with Ultralytics last May, through which it supports developers in implementing AI inference capabilities in robots, industrial equipment, intelligent cameras, and more by utilizing YOLO models.

In addition, we are expanding our collaboration with PaddlePaddle, an open-source deep learning framework with a developer base of over 4 million. DeepX explained that it has applied PaddlePaddle's OCR model, 'PP-OCR 5th Generation,' to its AI accelerator 'DX-M1' to support related AI models in operating in an NPU environment.

In terms of hardware, it is also pursuing integration with the Raspberry Pi ecosystem. DeepX announced that it launched an AI acceleration processor module for the Raspberry Pi 5 last June, providing a foundation for developers to test and verify AI applications and then expand them into industrial products.

DeepX presented a diffusion structure that connects the developer experience to industrial mass production as the core of this strategy.

The company plans to establish an environment that connects the AI model ecosystem with developer hardware, NPUs, software development kits (SDKs), and industrial references, enabling the entire process from development to proof-of-concept and commercial product development.

Kim Nok-won, CEO of DeepX, stated, “In the era of physical AI, an entire ecosystem is needed where developers can create models and deploy them to actual industrial sites,” adding, “We will pursue the establishment of an open execution platform that connects developers, AI models, and hardware.”