The first data from real homes.

Building the first motor priors datasets for physical AI. Training-ready, multimodal human-activity data, captured by everyday people in the homes they actually live in, not staged rooms. Made by international and national gold medalists in math and robotics.

  • NTUFounded by NTU Singapore robotics researchers
  • MARS LabNVIDIA-backed research partner
  • 88%Usable data rate

The world’s most diverse sample dataset

Request it free
50Hours of data
60+Tasks
20+Environments
2–5 minThen the task or the place changes

Every physical AI company needs a foundation model first.

Before any task-specific training, robots need motor priors, a general sense of how to reach, grip, and move through a real space. That’s 90–99% of their training data, and it has to come from real people.

open+ pour+ wipe+ fold+ carry+ … =

Real homes. No saturated tasks.

For the first time, capture isn’t happening in staged rooms and warehouses, with the same tasks repeated until the hours pile up. Models can now learn in homes people actually live in.

Typical capture dataset

A few tasks hold most of the hours.

Most recorded taskLeast
Eigen Robotics

Hours stay even across every task.

Most recorded taskLeast

Three streams. One moment.

Every session records video, hand motion, and 3D space at the same time, from the wearer’s eyes. All three stay in sync.

rgb
rgb
1 of 3

Visual

The task from eye level, exactly as it happened.

Egocentric RGB
rgb + hand + body
2 of 3

+ Motion

Every finger joint, grip, and reach, tracked through the whole task. The fine skill robots need to learn.

Hand pose, body pose
rgb + motion + spatial
3 of 3

+ Spatial

Room shape and object positions, mapped as the person moves. Same moment, same frame, every layer in sync.

Depth, SLAM, 3D scene

We build every part.

The hardware, the app, and the data engine are all built and run in-house. That’s how quality holds as the hours grow.

  1. Eigen Robotics wearable capture device with stereo cameras
    1Wear

    Custom hardware.

    Contributors wear our capture device and go about their day at home.

  2. Contributor app showing today’s recordings
    2Record

    Tasks with a purpose.

    Our app assigns the tasks, so every gap in the dataset gets filled on purpose.

  3. rawprocessed
    3Process

    Raw in. Data out.

    Every session is synced, privacy-blurred, and labeled.

  4. 4Train

    Foundation models.

    Your models learn how people move, from real people in real homes.

Quality you can measure.

88%

of captured data is usable. The rest is cut before it reaches you.

0100
4 months

from idea to production hardware and a sample dataset.

ideasample dataset

Our hardware is built for mass production, so hours grow with demand and quality holds.

Built by gold medalists and robotics researchers.

Our team includes international and national gold medalists in math and robotics. Eigen Robotics was started by researchers from Nanyang Technological University (NTU) in Singapore, and works with NTU’s MARS Lab on physical AI.

#1NTU in AI worldwideShanghaiRanking
Top 5NTU in data science and AIQS World Rankings
Research partner

NTU MARS Lab

NVIDIA-backed research in physical AI, multimodal LLMs, and robotics.

  • NVIDIA Academic Grant for multimodal robot manipulation, 2026
  • Papers at ICLR, CVPR, ICRA, and RSS
  • ICLR oral, top 1.8%
  • Meta Project Aria participant
  • ECCV challenge winner

MARS Lab collaborators

  • Sony
  • NVIDIA
  • Sharpa
  • MIT
  • ETH Zurich
  • Collaborator
  • Collaborator
  • HarvardHarvard
  • UC BerkeleyUC Berkeley
  • StanfordStanford
  • OxfordOxford
  • PrincetonPrinceton

Train on the real world.

Get the sample and see what real homes add to your model.

data@eigenrobotics.___