Real world first
We validate on hardware in messy environments, early and often — not only in simulation.
We think robots should leave the lab.
Most of what gets called AI never has to survive contact with gravity, friction, or a room it wasn't trained in. ET Robotics exists to build the part that does.
Building the physical layer of artificial intelligence.
Software intelligence has scaled faster than almost anyone predicted. Physical intelligence hasn't kept pace — most robots still fail the moment the world stops looking like their training data.
We design perception, control, and embodiment as one system, so machines can act competently in places nobody has fully mapped for them: warehouses, job sites, homes, and everywhere else "unstructured" actually means something.
Four ideas that shape every decision.
Full-stack thinking
Hardware, control and models are designed as one loop, never handed off between silos.
Safety by design
Compliant control and redundant safeguards are requirements, not features added later.
Ship deliberately
Fast where iteration is cheap. Slow where mistakes are expensive or dangerous.
Early, deliberate, and building alongside our first partners.
ET Robotics founded
Started on a straightforward bet: embodiment is the constraint holding physical AI back — not model quality alone.
Building the core stack
Perception, control and hardware platforms in active development, with early design partners feeding real constraints back into the work.
First real-world deployments
Moving from prototypes to systems that run for weeks at a time outside a controlled lab environment.
A small team on a hard, physical problem.
If that sounds like the kind of work you want to be doing, we'd like to hear from you.