Tau Robotics
About
About Tau Robotics
Tau brings a humanoid robot to a customer's property, where a trained operator controls it live from the company's facility. The cleaning tasks include wiping work surfaces, vacuuming floors, removing rubbish and moving ordinary lightweight items. A working network connection is part of the service, and an authorised adult remains at home throughout the visit.
The company records cleaning sessions and retains footage and telemetry for model training and development. Its July 2026 service terms describe one-hour home visits booked for a flat fee, with access limited to invited customers. This combines a paid cleaning workflow with collection of real household manipulation data, while human operators perform the current work.
Observed use
- DYNAMIXEL X Series at Koch v1.1 follower arm
2026-04-23 - Reference Design - Embedded - The Tau Robotics-derived Koch v1.1 follower arm uses DYNAMIXEL XL430 and XL330 servos in the ROBOTIS kit configuration.
Source
Team
- Alexander Koch - Co-Founder & CEO - Leads Tau Robotics and its robot-AI and commercial service programme.
- Cornelia Weinzierl - Co-Founder - Co-founded Tau Robotics and supports company operations.
Research
- Tau Robotics publishes world-model architecture for robot reinforcement learning
2025-05-20 - Research Publication - Tau Robotics published a technical report introducing its Latent Autoregressive Flow-Matching world-model architecture. The report describes a 1B-parameter model for predicting real-world image frames in compressed latent space. Tau positions the work as a path toward robot policies learned from large-scale real-world data with minimal human supervision.
Source
Timeline
Events
- 2025-05-20 - Research Publication - Tau Robotics publishes world-model architecture for robot reinforcement learning
Tau Robotics published a technical report introducing its Latent Autoregressive Flow-Matching world-model architecture. The report describes a 1B-parameter model for predicting real-world image frames in compressed latent space. Tau positions the work as a path toward robot policies learned from large-scale real-world data with minimal human supervision.
Source