Research

Our research teams investigate the foundations of embodied intelligence — from egocentric perception and world modeling to evaluation and real-world deployment — so that robotic systems can operate safely and effectively alongside people.

Egocentric Data

The Egocentric Data team focuses on collecting and curating first-person perspective data for robot learning. Starting with wearable sensor arrays, then extending to haptic gloves for tabletop manipulation, and integrating full-body exoskeleton systems for diverse environment data collection.

Robotics World Model

The World Model team builds dynamic world simulators that generate corresponding environments and robot actions from scene images and policy inputs. This enables human-in-the-loop interaction for post-training evaluation and robot policy refinement within a sandbox environment.

Evaluation & Post Training

The Evaluation & Post Training team develops systematic benchmarks and fine-tuning pipelines for robotic policies. We design human-in-the-loop evaluation protocols, build automated testing frameworks, and create post-training optimization methods that improve policy performance through iterative refinement.

Deployment

The Deployment team bridges the gap between research and real-world applications. We develop robust sim-to-real transfer methods, safety-critical control systems, and scalable deployment frameworks for industrial and service robotics.

Join the Research team

We're looking for researchers and engineers who are passionate about building the foundations of embodied intelligence.

See open roles