How Rhoda AI Plans to Bring Robots from Controlled Laboratory Demonstrations into Real-World Environments

Rhoda AI, redefining robotic intelligence, led by Jagdeep Singh, Eric Chan, Gordon Wetzstein, Andrew Wooten, Changan Chen, Steve Tirado, Alex Bergman, and Vincent Clerc, has announced its public launch after 18 months in stealth, unveiling FutureVision, a new approach to robotic intelligence based on video-predictive control and designed to operate beyond controlled laboratory demonstrations and into real-world environments. The company also announced it has raised $450 million in Series A funding to accelerate development and industrial deployment.

Traditional industrial robots excel in structured settings but struggle with real-world variability. Rhoda pre-trains models on hundreds of millions of videos to learn motion and interaction, then fine-tunes on smaller robot datasets to map predictions to actions.

The system continuously observes the environment, predicts future states as video, converts predictions into actions, executes them, and repeats this loop every few hundred milliseconds.

Rhoda’s proprietary Direct Video Action (DVA) model bridges perception and control, updating behavior dynamically as conditions change. The model’s strong motion prior from video-based pretraining enables efficient learning of new tasks, often with as little as ten hours of teleoperation data.

Built on this architecture, FutureVision acts as Rhoda’s intelligence layer, a foundation model that currently powers Rhoda systems and is expected to be licensed to partners across diverse robotic hardware and software platforms.

“The next era of robotics requires models that understand how the world moves, not just how it looks or is described in language,” said Jagdeep Singh, cofounder and CEO of Rhoda. “By learning from internet-scale video and operating in closed loop, these systems adapt to real-world variability in ways conventional approaches cannot. The goal is clear: robots that function effectively outside controlled lab settings.”

Rhoda’s technology has demonstrated autonomous operation in production environments with constantly changing materials, layouts, and workflows. In a recent high-volume manufacturing test, Rhoda completed a component-processing workflow in under two minutes per cycle without human intervention, surpassing customer KPIs.

“In manufacturing, high-variability tasks have resisted automation, with consistency under real-world conditions as the main challenge,” said Jens Wiese, Managing Partner at Leitmotif. “Rhoda’s approach adapts where humans are usually needed, expanding automation and supporting the re-industrialization of mature economies.”

The $450 million Series A will fund ongoing research and engineering, expansion of industrial deployments and customer pilots, and growth of Rhoda’s multidisciplinary team in generative AI, computer vision, and robotics.

“The first company to deploy intelligent, manipulation-capable robots at scale in real-world environments will generate a powerful data flywheel, creating a compounding advantage in capturing edge cases,” said Sandesh Patnam, Managing Partner at Premji Invest. “Rhoda has assembled the technical foundation, ambition, and execution capability to achieve this, making it an exceptional partner for advancing the next generation of intelligent robots.”

The company is supported by leading technology investors, including Capricorn Investment Group, Khosla Ventures, Leitmotif, Matter Venture Partners, Mayfield, Premji Invest, Prelude Ventures, Temasek, and Xora, alongside Silicon Valley figures such as John Doerr.

Rhoda is led by CEO and cofounder Jagdeep Singh, a serial deep-tech founder, and Chief Science Officer Eric Ryan Chan, a Stanford researcher and former generative model architect at WorldLabs. The leadership team also includes Gordon Wetzstein, Stanford professor and head of the Computational Imaging Lab, supported by experts from leading generative AI, computer vision, and robotics organizations.