Ropedia Raises $30M to Power Physical AI

The big picture: The race to build smarter robots has highlighted a critical data gap: physical AI cannot learn from internet data alone. Singapore startup Ropedia has developed a platform to collect and deliver real-world multimodal data, securing $30 million in pre-A funding to expand its operations.

Why it matters:

• Data bottleneck: Robotics companies need vast amounts of high-quality data showing human interaction with the physical world to move AI systems beyond labs into practical applications.
• Unique capture method: Ropedia’s HOMIE wearable device directly captures synchronized first-person video, audio, depth, gaze, hand tracking, and body movement, providing comprehensive human experience data.
• Cost efficiency: By collecting data from people rather than relying on expensive robot hardware, Ropedia claims to reduce data collection costs by up to 50 times compared to conventional approaches.

How it works:

• Multimodal data capture: The head-mounted HOMIE system records synchronized streams of various data types, all timestamped to teach AI models how perception and action occur together in real time.
• Processing and annotation: Raw synchronized recordings flow through Ropedia’s platform for processing and annotation, transforming them into model-ready training datasets for robotics and embodied AI developers.
• Scalable dataset: Ropedia has built Xperience-10M, a dataset with over 10 million interaction episodes and 10,000+ hours of multimodal recordings, continuously expanding with each new deployment.

The catch: While Ropedia addresses a critical need for physical AI data, the market for large-scale multimodal interaction data pipelines is nascent and could attract significant competition. Ensuring data quality, privacy, and the ability to generalize across diverse robotic platforms will be crucial for long-term success, as will fending off potential in-house data collection efforts from major robotics and AI companies.