Power to the Bot — CRV Partners With Dyna Robotics to Commercialize Embodied AI With Low-Cost Robots
By Max Gazor and Brian Zhan
Robots meticulously folding laundry. Robots preparing food with precision. These aren’t scenes from a sci-fi movie — Dyna Robotics is making them a reality. Today we’re thrilled to announce the $23.5 million seed round for Dyna Robotics, co-led by CRV.
Different Perspectives
Unlike the flashy, one-size-fits-all visions of traditional robotics, the team at Dyna Robotics is laser-focused on building practical, cost-effective solutions that work in real production environments. While much of the robotics buzz has centered on high-cost humanoids or complex, all-purpose machines, Dyna Robotics is taking a radically different approach: start simple and excel. Their robots are designed to master one task at a time — whether it’s folding garments or preparing food — using an affordable pair of stationary robotic arms. They’re taking a practical approach of leveraging task-specific data to build embodied AI that learns and improves with every action.
Pioneering Founders
At the heart of Dyna Robotics are three visionary founders whose backgrounds span from breakthrough research to successful entrepreneurial exits. Co-founders Lindon Gao and York Yang are no strangers to transforming industries — they previously built and exited Caper AI to Instacart for $350 million, revolutionizing retail with smart carts that now serve grocery stores worldwide.
Joining them is Jason Ma, a former DeepMind Robotics and NVIDIA Robotics research scientist. Jason Ma’s contributions to robotics have reshaped how we think about robot learning. Unlike traditional methods which rely on human-engineered rewards, Jason and the NVIDIA team brought a new AI agent called Eureka to life. Eureka allows robots to create and refine their own reward functions, dramatically improving their ability to master complex tasks. This was demonstrated by Jensen Huang at NVIDIA GTC last year, where a simulated Shadow Hand robot was shown performing pen spinning with unprecedented skill — an achievement that would have been nearly impossible with traditional approaches. His subsequent work, like DrEureka, that taught a robot dog how to walk and balance on a yoga ball in the real world, broke new grounds in foundation models for robotics.
The Opportunity
Why now? Despite the meteoric rise of language, image and video models, embodied AI has remained elusive. Unlike digital domains where vast amounts of data are just a click away, the physical world presents a scarcity of high-fidelity, task-specific data. Dyna Robotics recognized that to build robots that learn effectively, they needed to generate their own extensive datasets — capturing everything from the art of packaging to the meticulous act of cleaning.
Across sizes and industries, companies have tasks they would readily offload to robots if the right solutions existed at the right price points. Despite thriving language, image and video models, embodied AI still hasn’t cracked the code. By focusing on tasks achievable with simple yet effective hardware, the company is not only making robotic solutions accessible, but also setting the stage for continuous, real-world learning.
Once in a blue moon, we hear the vision of a founder where the ambition and potential scale of impact is so large that we decide to partner together in the same week which was the case with the Dyna Robotics team. We remember the awe we felt as we imagined how the world would change if they succeeded… and those feelings have only grown stronger when we saw the robots in action.
This isn’t just incremental progress. It’s a paradigm shift in how we think about robotic intelligence. Instead of relying solely on simulations, Dyna Robotics is grounding its models in the messy, unpredictable real world — ensuring that their solutions are both robust and economically viable. Here at CRV we’ve been backing innovative founders since 1970. If you’re working on exciting groundbreaking technology our team would love to hear from you.
