Rhoda AI Launches Publicly After Stealth With $450M Funding and New AI Robotics Platform
After operating quietly for more than a year and a half, Rhoda AI has officially emerged from stealth mode and announced its public debut along with a major funding milestone. The robotics and artificial intelligence company revealed that it has secured $450 million in Series A financing while introducing a new approach to robotic intelligence designed to bring advanced robots out of controlled laboratory environments and into real-world industries.
At the center of the company’s technology is FutureVision, a robotics intelligence system based on video-predictive control. The platform represents Rhoda’s effort to solve one of the most persistent challenges in robotics: enabling machines to operate reliably in complex, unpredictable environments.
With the new funding, the company plans to accelerate research and development, expand industrial deployments, and grow its multidisciplinary engineering team across artificial intelligence, computer vision, and robotics.
Addressing the Limits of Traditional Robotics
Industrial robots have long played an important role in manufacturing, logistics, and other industrial sectors. However, most traditional robotic systems rely heavily on pre-programmed movements and operate best in structured environments where every object and motion path is carefully defined.
These robots typically follow fixed trajectories, meaning they are highly efficient when performing repetitive tasks in stable conditions but struggle when faced with variability. Small changes in object positions, lighting conditions, or workflow sequences can disrupt their operations.
Over the past decade, researchers have explored new AI-based methods to make robots more adaptable. One promising approach involves vision-language-action (VLA) models, which allow robots to interpret visual information, understand instructions, and perform actions based on learned patterns.
While these systems have demonstrated impressive capabilities in research environments, many still encounter difficulties when transitioning to real-world settings. Real environments often involve unexpected obstacles, irregular materials, dynamic layouts, and countless edge cases that traditional AI models may not be trained to handle.
Rhoda AI was founded to address this gap between laboratory performance and real-world deployment.
Learning From Internet-Scale Video Data
Rather than relying primarily on traditional robot training methods, Rhoda AI uses a different strategy to teach machines how to interact with the physical world.
Instead of collecting massive datasets of robot demonstrations through teleoperation, the company pre-trains its models using vast amounts of publicly available video data. This dataset includes hundreds of millions of videos sourced from the internet.
By analyzing this large volume of visual data, the system learns fundamental patterns about how objects move, how forces interact, and how physical environments behave. The training process allows the model to build a strong “motion prior,” meaning it develops an intuitive understanding of physical interactions and movement dynamics.
After this large-scale video pretraining stage, the models undergo a second phase of training using smaller datasets generated from real robot interactions. This step teaches the system how to translate visual predictions into specific robotic actions tailored to particular hardware systems.
This combination of large-scale visual learning and targeted robotic training enables the models to adapt quickly to new tasks.
Continuous Real-Time Interaction With the Environment
Rhoda’s robotic systems operate using a continuous feedback loop that allows them to observe, predict, act, and adjust in real time.
The process begins when the robot observes its environment through sensors and cameras. The system then predicts possible future states of the environment as a sequence of video frames. These predicted outcomes are translated into robot actions, which are executed immediately.
Once an action is completed, the robot observes the environment again and updates its predictions accordingly. This cycle repeats every few hundred milliseconds, allowing the robot to adjust its behavior continuously as conditions change.
This approach enables robots to respond dynamically to unexpected situations, making them far more adaptable than systems that rely solely on pre-defined plans.
The Direct Video Action Model
The architecture powering Rhoda’s technology is known as the Direct Video Action (DVA) model.
Unlike traditional robotic control systems that generate action plans in advance, the DVA model integrates perception and control into a unified framework. The model continuously updates its predictions and actions based on new sensory input.
This closed-loop design allows robots to respond to environmental changes in real time while maintaining awareness of the physics and motion patterns within their surroundings.
Because the system is trained on extensive video data, it also learns strong motion dynamics during the pretraining phase. This capability enables the model to acquire new skills quickly with relatively small amounts of robot-specific training data.
According to Rhoda, some tasks can be learned with as little as ten hours of teleoperation data, dramatically reducing the time required to train new robotic capabilities.
FutureVision as a Robotics Intelligence Layer
Built on the Direct Video Action architecture, FutureVision serves as the core intelligence layer powering Rhoda’s robotic systems.
Rather than being tied to a single type of robot hardware, the platform is designed as a foundation model that can potentially be applied across different robotic platforms and industries.
In addition to supporting Rhoda’s own robotic deployments, the company expects that FutureVision may eventually be licensed to partners developing their own robotic hardware and software ecosystems.
This model-based approach mirrors trends seen in other areas of artificial intelligence, where foundation models are used as flexible building blocks that can be adapted to multiple applications.
Early Industrial Deployments
Although the company has only recently emerged from stealth mode, Rhoda reports that its technology has already been tested in real industrial environments.
In manufacturing scenarios, robots must often handle a wide range of materials, components, and production workflows. These environments frequently involve unpredictable variations that have historically been difficult to automate.
During a recent high-volume manufacturing evaluation, Rhoda’s system successfully completed a component-processing workflow in less than two minutes per cycle without requiring human intervention. The performance exceeded the operational targets established by the customer.
Such demonstrations highlight the potential for adaptive robotic systems to automate tasks that were previously considered too complex or variable for traditional automation solutions.
Expanding Automation in Manufacturing
Industry experts believe that technologies capable of handling variability could dramatically expand the range of tasks that robots can perform.
Jens Wiese, managing partner at venture capital firm Leitmotif and a former executive at Volkswagen Group, noted that high-variability tasks have historically been difficult to automate.
According to Wiese, the real challenge in industrial automation is not simply solving a task once but ensuring that the solution can deliver consistent performance across changing production conditions.
He suggested that technologies capable of adapting to dynamic environments could significantly expand automation opportunities in manufacturing, logistics, and other industries.
Such advancements may also contribute to broader economic trends such as the revitalization of manufacturing capabilities in developed economies.
Major Investment to Accelerate Development
Rhoda’s $450 million Series A funding round will support continued development of the company’s robotics technologies.
The funding will be used to expand engineering and research efforts, develop additional capabilities for the FutureVision platform, and support new industrial deployments and pilot programs.
The investment also reflects strong interest from major technology investors who believe that intelligent robotics represents one of the next major frontiers in artificial intelligence.
Sandesh Patnam, managing partner at Premji Invest, emphasized the strategic importance of deploying real-world robotic systems at scale.
He explained that the first company capable of deploying intelligent robots across diverse real-world environments could establish a powerful data advantage. As robots gather more operational data, the resulting feedback loop would allow the company to continuously improve its models and capabilities.
Backing From Leading Technology Investors
Rhoda AI has attracted support from several prominent venture capital firms and institutional investors. Participants in the funding round include Capricorn Investment Group, Khosla Ventures, Leitmotif, Matter Venture Partners, Mayfield, Premji Invest, Prelude Ventures, Temasek, and Xora.
The company has also received backing from prominent technology leaders in Silicon Valley, including John Doerr.
Leadership and Research Expertise
Rhoda AI is led by cofounder and CEO Jagdeep Singh, a serial deep-technology founder who has built and scaled multiple companies in advanced technology sectors.
The company’s scientific leadership includes Eric Ryan Chan, who serves as chief science officer and previously worked as a generative model architect at WorldLabs.
Another key member of the team is Gordon Wetzstein, a professor at Stanford University and head of the Computational Imaging Lab.
Together with a multidisciplinary team drawn from leading AI, robotics, and computer vision organizations, the leadership group is focused on building systems capable of operating reliably in real-world environments.
Toward Real-World Intelligent Robots
The emergence of Rhoda AI reflects a broader shift in the robotics industry toward systems capable of learning and adapting in dynamic environments.
While many robotic systems have traditionally been limited to controlled factory settings, new AI-driven approaches are opening the door to more flexible and autonomous machines.
By combining large-scale video learning, real-time prediction, and closed-loop control, Rhoda aims to create robots that can operate effectively outside the lab.
If successful, technologies like FutureVision could transform industries ranging from manufacturing and logistics to construction and agriculture—bringing the vision of truly intelligent robots closer to everyday reality.
Source link: https://www.businesswire.com
