Jiben Intelligence Completes Two Consecutive Funding Rounds, Raising Nearly Ten Million USD in Total
Jiben Intelligence recently completed two consecutive funding rounds totaling nearly ten million USD, led respectively by Alphaist Partners and Glory Ventures, with follow-on investment from Jieli Angel and the actual controller of a certain components industry player. Founded in September 2026 by two post-2000s researchers, the company focuses on dexterous robotic manipulation, takes a reinforcement learning approach, and is exploring the creation of a manipulation model platform adaptable to different dexterous hand hardware.
Body · AI translation
It is reported that Jiben Intelligence has recently completed two consecutive rounds of financing, with a cumulative amount of nearly ten million US dollars. The two rounds were led by Alphaist Partners and Yaotu Capital respectively, with Jieli Angel and the actual controller of a certain components industry player participating as follow-on investors. Jiben Intelligence was founded in September 2026 by two post-2000s researchers. CEO Li Puhao is studying in the Department of Automation at Tsinghua University and has conducted research on dexterous hands and dexterous manipulation since 2021. CTO Chen Zeyuan holds a master's degree from Peking University, has long researched underlying interaction algorithms for dexterous manipulation, participated in work such as ClutterDexGrasp and CADGrasp, and has conducted research at Amazon FAR, Zhiyuan and JD.com. In terms of core strategy, Jiben Intelligence mainly focuses on robotic dexterous manipulation, committed to building a software-hardware integrated dexterous hand solution, with algorithms, data and engineering capabilities as the core, to enhance robots' fine manipulation capabilities in real environments. Specifically, Jiben Intelligence has chosen the reinforcement learning route, enabling models to learn manipulation strategies through interaction, trial and error, and feedback. Currently, the company is exploring the combination of simulation, human demonstrations and generated interaction trajectories to build a training data system, and is advancing the validation of model transfer from simulated environments to real hardware. Jiben Intelligence hopes to build a manipulation model platform that can adapt to different dexterous hand hardware: upward, it receives manipulation goals given by multimodal large models; downward, it adapts to different hardware, providing callable manipulation capabilities for robotic systems. While advancing model research and development, the team is also considering accelerating technical validation and accumulating data and feedback from real scenarios through self-developed hardware or cooperation with hardware manufacturers.