PsiBot develops AI models and dexterous manipulation hardware for robotics. Their technology combines dual arms, five finger dexterous hands, and wheeled chassis for embodied AI. Founded by veterans from JD.com Robotics, Huawei, Tencent, and Stanford AI Lab.
PsiBot is a robotics startup developing AI and hardware for dexterous robotic manipulation. Founded in 2024 by product and research leaders from JD.com Robotics, Huawei, Tencent, and Stanford-linked laboratories, the company focuses on enabling robots to perform complex physical tasks through the integration of advanced manipulation hardware and reinforcement learning. Its robotic platforms combine dual-arm manipulation, five-finger dexterous hands, and mobile bases with AI-driven control systems designed for general-purpose manipulation. By concentrating on one of the most challenging problems in robotics, PsiBot aims to give robots the dexterity and adaptability needed to operate across diverse industrial and commercial environments.
In 2025, PsiBot highlighted recognition for its reinforcement-learning leadership, including Forbes Asia 30 Under 30 for co-founder Yuanpei Chen. The company used that moment to reinforce its message that simulation-to-reality dexterous manipulation was its central technical moat.
PsiBot was founded in 2024 by a team spanning robotics productization, reinforcement learning, simulation, and full-stack engineering. The company positioned itself around the challenge of combining high dexterity, generalization, and success rate in embodied AI systems.
In November 2024, PsiBot announced an angel round led by GL Ventures and Lanchi Ventures. It said the funding would support reinforcement-learning-based robot skills, scenario-driven data generation, and end-to-end dexterous manipulation systems.
In 2025, PsiBot highlighted recognition for its reinforcement-learning leadership, including Forbes Asia 30 Under 30 for co-founder Yuanpei Chen. The company used that moment to reinforce its message that simulation-to-reality dexterous manipulation was its central technical moat.
PsiBot is approaching embodied AI from the manipulation bottleneck rather than from locomotion alone. Its thesis is that progress in reinforcement learning, simulation, and data generation can unlock general-purpose dexterous manipulation agents that are directly useful in commercial environments.
