We are a research company building open, physical, and planetary‑scale AI — the foundational infrastructure for humanity’s leap outward. Models that think, machines that move, systems that endure.
From the model layer to the metal, we build the full stack of embodied, open intelligence.
Foundational models, training frameworks, and tooling — built in the open so the whole field can stand on them.
Perception and control stacks for drones that navigate, cooperate, and reason in the open sky.
Embodied agents that sense, grasp, and act — bridging the gap between language and the physical world.
Reliable, auditable AI for the businesses that run the world — deployed on your terms, your data, your metal.
Fundamental work on reasoning, world models, and scaling — publishing what we learn as we learn it.
Interpretability and control research woven through everything — because power without wisdom is a hazard.
Language models learned to think. Now they learn to move. Celestris couples frontier reasoning with real-time control, so machines can perceive a changing world and act inside it — on factory floors, over farmland, and one day, off-world.
We believe the safest way to build powerful intelligence is transparently — releasing our infrastructure, publishing our research, and weaving interpretability and control into every system before it ever touches the world.
We’re hiring researchers, roboticists, and engineers who want to reach beyond the horizon. And we’re opening early access to our infrastructure.
Tell us who you are and what you want to build. We read every message and reply personally.