The Data Infrastructure Layer for Robotics
Data collection hardware, an active data pipeline with a growing dataset, and a simulation environment for foundational model training.
_
Robotics foundation models need millions of hours of human demonstration data. Existing capture systems are either prohibitively expensive, proprietary, or limited to arm-only tracking. Eidon built the full stack from scratch.
Eidon Tracker
7-IMU wearable capturing full upper-body arm kinematics — shoulder, elbow, forearm, and wrist — at 24Hz over BLE. 8–12hr battery life. Field-deployable. Over 1,200 hours already collected.
Eidon Glove
16-DOF finger tracking via Hall effect sensors and embedded magnets. Per-joint angles at every segment of every finger. ~100Hz over BLE. Compatible with MuJoCo and Isaac Gym out of the box.
Eidon App ↗
iOS and Android. Syncs natively with trackers for synchronized video + sensor data collection. Video-only mode, integrated QC pipeline, and built-in operator payment handling.
Eidon Sym ↗
Simulation environment and data pipeline. VLM-powered quality control runs daily. Outputs simulation-compatible formats. 3D visualization, manual labeling, and trajectory playback synced with video.
_
Available on iOS and Android. The Eidon App syncs natively with our trackers and glove to capture synchronized egocentric video and sensor data in the field. It also supports video-only collection mode and handles operator payments natively.
Equip and connect
Operators wear the Eidon Tracker and Glove, mount their phone, and open the app. The hardware pairs automatically over BLE — no calibration facility or lab setup required.
Capture synchronized data
The app records egocentric video alongside full arm kinematics and finger articulation data in a single synchronized session. Operators perform everyday manipulation tasks from their own environment.
Automated QC and delivery
Collected data flows into Eidon Sym, where a VLM-powered pipeline runs daily quality control — filtering, auto-tagging objects, and outputting simulation-compatible formats ready for model training.
