The Skills Layer for Physical AI

We license egocentric video and sensor data.

We are looking for approximately 500,000 hours of highly-varied egocentric and teleops data — video with synchronized sensors, grippers, dexterous hands and the like. We license off-the-shelf footage of people doing real tasks with their hands in frame, narrating as they go. And we commission bespoke capture programs that specify the modalities, devices, settings and environments up front. Tell us what you have and we will read it for fit.

What we look forTask-segmented, first-person, real environments
What matters mostClear rights and signed consent on every clip
TurnaroundA read on fit within three business days
Off the shelf

Tell us what you have off the shelf.

01

About you

02

Your off-the-shelf library

Rough numbers are fine. We are sizing the opportunity, not auditing you.

03

Capture device mix

Share of total hours by device. Leave a row at zero if it does not apply.

Device total0%
04

Sensor and modality mix

Hours carrying each signal alongside the video.

Example clip record — the standard we license against

clip_id
EGO-2026-08841
duration
00:04:12
task_label
load_dishwasher / mixed_load
segmentation
task-segmented, one task per clip
capture_device
GoPro Hero 12, head mount
resolution / fps
5312 x 2988 @ 60
modalities
RGB, IMU, audio
depth
none
operator_id
hashed
consent
signed release on file
rights
AI training permitted, perpetual
environment
residential kitchen
05

Specialty equipment

Describe the capabilities of any specialty cameras or sensors in the library.

We reply to every submission. If the library is a fit we will send a mutual NDA and a sample request.

© 2026 Ergodexa, Inc. hello@ergodexa.com