Former Meituan Waimai Head Enters Embodied AI for Restaurants, AtomBite.AI Raises Tens of Millions in Seed Funding

AtomBite.AI (Yuanjie Intelligent), an embodied AI startup, has completed a multi-million yuan seed funding round led by Inno Venture Capital, with participation from Tsinghua Alumni Seed Fund and other angel investors. The company is founded by Dr. Wang Dong, former technical head of Meituan's food delivery unit, who led a team of over 1,000 engineers and built the algorithms, data, and system architecture supporting tens of millions of daily orders.

Unlike many startups chasing general-purpose humanoid robots, AtomBite.AI is laser-focused on restaurant back kitchens—a domain with acute, global pain points driven by rising food delivery orders and structural labor shortages. "Restaurants don't care if your robot looks human," Wang says. "They care about cost, capability, and cost savings."

The company's core technology is a "World Action Model" (WAM), inspired by Yann LeCun's vision of predictive world models. Instead of reactive "see-and-do" approaches (VLA models), WAM enables robots to simulate the outcomes of actions before moving. This is critical for chaotic kitchen environments with smoke, occlusion, and random object placements.

AtomBite.AI's technical stack has three layers: a top-level embodied world model that understands kitchen physics and predicts action consequences; a mid-level task scheduling and optimization engine that translates decisions into precise physical plans; and a bottom layer of custom key components integrated with general hardware for robust real-world operation.

The company's go-to-market strategy starts with food packaging and delivery handoff, the most error-prone link in the delivery chain. From there, it will expand to controlling existing kitchen equipment like fryers and dishwashers, ultimately aiming to build a "digital twin kitchen operating system" for global intelligent scheduling.

Dr. Wang, a PhD student of Academician Zhang Bo at Tsinghua, also built the world's first commercial video face recognition and tracking system in 2011. Co-founders include Li Tao, former head of Meituan Waimai's algorithm and data systems, and Li Haozhe, a serial entrepreneur with global business experience. Other team members come from Tsinghua, USTC, Meituan, and Horizon Robotics.

"2026 is the first year of embodied application," Wang says. "Only by forming a data-model-hardware flywheel in real environments can embodied AI move from lab to industry." The seed round marks just the beginning for the fledgling company, which has already secured product deployment intentions from top domestic and international companies.