Pre-seed · Now raising

The data engine for embodied AI.

Robots will learn from the real world before they master it. We are building the scalable, real-world data pipeline that the next generation of humanoid and embodied AI will be trained on.

Request our deck How it works

Egocentric: First-person task capture
Multimodal: Vision · motion · context
Scalable: Distributed contributors

Built for

Research labsApplied AIHumanoid roboticsVertical service robots

/ Thesis

Embodied AI will not be limited by models. It will be limited by data.

Self-driving reached production scale through massive real-world data loops. The home is messier than the road, tasks are open-ended, and success depends on real interaction — not passive observation. The companies that own the upstream data flywheel will become indispensable to the robotics ecosystem, regardless of who ultimately builds the hardware. That is the layer we are building.

/ Platform

Built for embodied AI — everything you need to pre-train, fine-tune, and benchmark humanoid robot models.

Egocentric capture

First-person recordings of real human activity — captured on our proprietary, custom-designed wearable kit and shipped to a distributed network of contributors.

Multi-modal streams

Synchronized video, IMU, depth, and audio for comprehensive scene understanding — the signal density embodied models actually need.

Rich annotations

Dense labels including hand poses, gaze direction, object interactions, and temporal action segments — ready for fine-tuning and benchmarks.

Pre-training ready

Large-scale, diverse datasets and continuously refreshed feeds — delivered via bulk API with versioning to robotics labs and humanoid programs.

Fine-tuning & benchmarks

Task-specific subsets with dense annotations for fine-tuning and standardized model evaluation.

/ Roadmap

The capability ladder.

Customers can buy data long before humanoids hit full autonomy. Each stage compounds the next.

01
Narrow tasks
High-quality data on focused household routines.
Stage 1
02
Broader skills
Cross-task generalization across environments.
Stage 2
03
Compositional
Sequenced, multi-step real-world execution.
Stage 3
04
General autonomy
Adaptive performance across diverse homes.
Stage 4

/ Collectors

Get paid to teach robots how the real world works.

Our distributed contributor network captures the everyday tasks that humanoids need to learn. We ship you our proprietary, custom-designed recording kit — contribute on your own schedule and earn for accepted submissions.

Become a data collector Apply now

Flexible

Contribute on your own time, from anywhere.

Paid

Transparent rates per accepted submission.

Private

Consent-first capture and review.

Meaningful

Your data trains real embodied AI.

/ Now raising

Help us build the data layer of embodied AI.

We're talking with mission-aligned investors, robotics partners, and early advisors. If that's you, we'd love to share more.

Get in touch