Real-world robot learning

Data and infrastructure for autonomous robotics

Multi-modal Data

Synchronized video, motion, depth, and poses from real-world activity.

Research-driven Annotations

State-of-the-art annotation precision, backed by Trace Research and benchmarked against ground truth.

Tooling for Research

Browse, query, and evaluate datasets built for how robot-learning teams actually work.

Diverse Tasks & Environments

Across thousands of real homes, workplaces, and tasks we are building the library of real-world physical work.

WHY TRACE

The real world,
made legible.

Robots learn from experience, and the richest source of it is everyday activity and work. Trace captures that activity and turns it into structured, high-fidelity data.

How we do it

We trace,
robots follow.

Every task we capture – a hammer swing, a poured cup, a threaded needle – becomes a structured, multi-layered record: the motion, the contact, the intent. We don't just record the physical world. We make it legible to machines.

Trusted by teams building embodied AI

Powering the open source research behind robot learning

the egoverse consortium

About the EgoVerse

Join us.

Explore Careers with Trace

Real-world robot learning

The real world, made legible.

We trace,robots follow.

Join us.

Building the interface between robotics research and the physical world.

The real world,
made legible.

We trace,
robots follow.