Ropedia raises $22M to scale data collection for training robots

Ropedia has raised $22 million in pre-Series A funding, bringing its total funding, including its seed round, to $30 million. The company, founded in 2025 and based in Singapore and Mountain View, California, plans to scale HOMIE, a wearable head-mounted system for collecting data to train physical AI.

HOMIE records first-person human movements, object interactions, and spatial context, sending the results to Ropedia’s processing and annotation models. The device has a 360-degree camera view, audio sensing that can reconstruct an audio source and direction, and a motion unit for tracking wearer movement. Ropedia said it develops much of the device’s overall hardware structure, circuit-board assembly, and sensor synchronization in-house.

The company’s pipeline runs from data capture through model-aligned fine-tuning. Unlike providers that label client-owned data, Ropedia creates and structures its own datasets. It also contrasts its method with teleoperation-based collection, which needs robot hardware and is generally tied to particular robot embodiments. Ropedia said its approach can reduce collection costs by as much as 50 times compared with traditional methods.

Ropedia is initially collecting data in homes for activities including housekeeping, cleaning, and interior design, and has also recorded pickleball practice through a club whose members wear its headsets. It currently collects about 1,000 hours of data weekly and has accumulated 100,000 hours. It licenses off-the-shelf datasets and acquires high-quality data for customers.

The funding will support teams in hardware, software, data infrastructure, and model development; expansion in North America, particularly the U.S.; platform features for annotation, quality analytics, and compliance; AI research; Southeast Asian collection; U.S. hiring; and increased HOMIE manufacturing. Planned hardware includes hand-tracking and tactile devices synchronized with HOMIE, plus shoes that sense foot-ground interaction.