Lili-o operates the world’s first fully autonomous robo‑centric data foundry, deploying robots in real homes to capture high‑fidelity, contact‑rich manipulation data tailored to household environments.
Funding
Funding not disclosed
Founders
Product
Problem
Robotics developers lack large-scale, high-fidelity manipulation datasets captured in real household settings, because existing data sources rely on expensive teleoperation rigs or low‑quality video streams that do not provide precise contact information.
Solution
Lili‑o operates an autonomous robo‑centric data foundry that deploys robots directly in homes to record contact‑rich manipulation interactions. A deterministic one‑shot execution engine captures a single demonstration and then generates repeatable, pose‑aware skills without additional training or cloud processing. The system produces cross‑embodiment datasets that reflect the variability of real‑world household environments, enabling robot teams to accelerate manipulation capability development. By turning manual data‑engineering into an around‑the‑clock factory, Lili‑o supplies robot manufacturers and research labs with ready‑to‑use training data that would otherwise require costly teleoperation setups.
Target Audience
Primary customers are robotics companies and research laboratories developing household manipulation capabilities for service robots.
Features
- Autonomous robots operate in actual residential spaces, capturing high‑precision force and pose data during manipulation tasks
- One‑shot imitation learning engine records a single demonstration and synthesizes repeatable skills that adapt to new object poses and clutter
- Cross‑embodiment data generation ensures compatibility with multiple robot platforms and hardware configurations
- Continuous, around‑the‑clock data collection eliminates the 1:1 human‑to‑robot operational bottleneck of teleoperation rigs
- Dataset includes rich contact modalities (force, tactile, joint trajectories) unavailable from video‑only or ego‑centric sources