Persona AI (CEO Yoo Seung-jae, hereinafter Persona AI) announced that it is accelerating the era of Physical AI by leading the development of a robot operating system (OS) based on its proprietary AI technology.

Persona AI has demonstrated its global technological competitiveness through a lightweight AI model that operates without internet connectivity or GPUs, earning back-to-back CES 2025 and 2026 Innovation Awards and winning international competitions such as the Gen AI Competition.

Persona AI’s proprietary AI engine integrates an on-device next-generation voice AI model, SSTT (Sovereign AI Speech to Text), which recognizes even dialects, and a language model, sSLM (Sovereign AI Small Language Model), which addresses issues of historical distortion. This enables ultra-fast processing, high security, and zero cloud costs. On this basis, the company has succeeded in realizing “agentic robots” that observe, judge, and act solely through voice commands, going beyond joystick control and fixed-scenario constraints of existing robots.

In April this year, Persona AI officially opened its Physical AI Lab and has begun full-scale Korea–U.S.–China joint research with U.S. university research teams, Chinese humanoid company Leju Robot and TI5, among others. The company also stated that it has integrated Persona AI’s voice AI into Stanford University’s Mobile ALOHA robot, significantly enhancing imitation learning and voice control performance.

The company plans to preempt the agentic robot market by optimizing its proprietary AI models, including LAM, VLA, and sLLM, for robot OS. The VLA (Vision-Language-Action) model is a core technology that enables robots to perform tasks on the spot by combining scenes captured by their cameras with users’ natural language commands.

Persona AI intends to prioritize application of its Korean-specialized STT to non-structured environments in defense, manufacturing, logistics, and household robots, enabling robots to operate without latency. It is also considering supplying its proprietary sLLM in a commercial SDK format by deploying it on-device to reduce cloud dependence and secure advantages in cost and security. Furthermore, the company aims to secure a unique position in the global market by acquiring next-generation Physical AI technologies, including joint training with NVIDIA Cosmos-type world models and expanded precise manipulation through the integration of tactile and force sensors.

A Persona AI representative stated, “We have unveiled a Physical AI robot based on our strengths in AI lightweighting and engine technology, and through this, we intend to lead the Physical AI ecosystem going forward.”

Popular News