Striding AI, a new player in the robotics and artificial intelligence landscape, has emerged from stealth today with ambitious plans to build foundational technologies for Physical AI. The company aims to enable robots to perceive, reason, act, and continuously improve through interaction with the physical world, addressing a critical need for scalable and intelligent robotic applications. This launch signifies a strategic step towards bridging the gap between advanced AI models and their practical implementation in diverse real-world settings.

The startup's approach integrates advanced foundation models with robotic perception, control systems, real-world action data, and robust deployment infrastructure. According to Song Yao, founder and CEO of Striding AI, breakthroughs in Physical AI stem from the continuous co-evolution of data, models, and infrastructure. The company's leadership team brings extensive experience from AI chips, autonomous driving, robotics research, and industrial technology, positioning them to bring complex technologies into production environments.

Striding AI plans to initiate its deployment efforts in structured environments, particularly within the retail sector. Here, robots will support tasks such as shelf restocking, inventory counting, product organization, and checkout assistance. These initial applications are chosen for their frequent human interaction, repeatable workflows, and rich operational data, providing an ideal starting point for developing scalable Physical AI systems. The long-term vision extends beyond retail, with expectations to expand into sectors including food, agriculture, logistics, healthcare, and telecommunications.

Early internal testing has already demonstrated significant promise, with Striding AI's human-in-the-loop reinforcement learning (RL) method improving task success rates by up to three times. To scale this efficiency, the company is actively building infrastructure for robot pretraining, distributed reinforcement learning, and edge-to-cloud orchestration. This comprehensive platform is designed to continuously improve as more robots operate and generate data in real-world scenarios, fostering a self-improving ecosystem for intelligent machines.