Yuanli Lingji Intelligent Technology

About Yuanli Lingji Intelligent Technology

Yuanli Lingji Intelligent Technology, operating as Dexmal, is a robotics and artificial intelligence firm developing embodied intelligence solutions. The company was established in March 2025 by a core team from AI company Megvii Technology, including Fan Haoqiang, Zhou Erjin, and Wang Tiancai. The founding team possesses strong academic backgrounds and extensive experience in AI-native product development, robotics hardware and software, and foundation model training.

Dexmal focuses on creating a comprehensive embodied intelligence technology stack that integrates large-scale models with robotic hardware to achieve end-to-end intelligent control. Its product ecosystem includes a self-developed end-to-end multi-modal embodied intelligence large model called MMLA, which integrates multi-sensor data with large language models (LLM), vision-language models (VLM), and vision-language-action models (VLA) to enable generalization across various scenarios and tasks. The company has also developed an open-source, PyTorch-based VLA toolbox named "Dexbotic" and an open-source robot hardware platform, DOS-W1, to facilitate easier deployment and maintenance. To advance research and development, Dexmal co-initiated RoboChallenge, a large-scale, real-robot-based VLA online benchmark, in collaboration with Hugging Face.

The company's business model centers on the research, development, and eventual commercialization of its software and hardware technologies for intelligent robots. Dexmal has attracted significant investment to fuel its growth. In March 2025, it secured a 200 million yuan angel round from investors like Legend Capital, Jiukun Venture Capital, and Qiming Venture Partners. This was followed by a Series A round led by NIO Capital and a Series A+ round led solely by Alibaba, bringing the total financing across these rounds to nearly 1 billion yuan (approximately $140 million). These funds are allocated to accelerate the transition from research to scaled application, standardizing data collection, training, and on-device inference to prepare for real-world pilots.