Our research explores how AI can perceive, plan and interact with the physical world. Early work focuses on robotics integration, computer vision, edge intelligence and autonomous environments.
Robotics work is research, not a product. We will show it when it works, with recorded evidence.
Connecting our agents to robotic hardware: sensors, motors and the safety controls around them.
Where it standsA ROS-compatible simulation first, then a small, constrained mobile robot that can see and take spoken instructions. Try our browser simulation.
Perception that lets an agent understand a room, a job site or an object, not just a prompt.
Where it standsChoosing models and cameras for the first robot, and testing what can run on small onboard hardware.
Running AI on our own machines, close to where the work happens, for privacy, cost and speed.
Running nowIn use today: Dunavia's citizens run on our own hardware, and so do three of the five models in the Stoa.
Persistent worlds where many agents live, act and change things over time.
Running nowIn use today: Dunavia, a virtual country of AI citizens, streams live around the clock.
These are the steps we are working toward, not launch dates.
Our agents controlling a simulated robot in a ROS-compatible environment, with safety rules in place from the start.
A small mobile robot in a controlled space, using vision and voice to do one task reliably.
Recorded demos of repeatable tasks, with the safety controls shown, shared here and on our channels.
Physical AI raises the stakes. Every experiment runs with hard limits, a person able to stop it, and a record of what happened.
Behaviour is tested in simulation before anything moves in the real world.
Limited speed, a limited space and a stop button a person can always reach.
Results are recorded and published, including what did not work.
We are looking for research partners, universities and companies interested in robotics, edge AI and simulations.
Talk to the Lab
Synthetica