We pursue the hard, foundational problems across six disciplines — and hold every result to a rigorous, reproducible standard.
A shared method holds the disciplines together. It is deliberately unglamorous — and it is why the work endures.
We reason from the physics up. Assumptions are stated, tested, and discarded when the evidence disagrees.
Every decision is made deliberately. If a detail does not serve the work, it does not ship.
We engineer for the long term. Longevity is treated as a specification, not an afterthought.
We chase hard questions, then hold the answers to rigorous, reproducible standards.
We study learning and reasoning as engineering problems: how to build systems that are capable, understood, and controllable. Alignment is not a downstream concern — it is designed in from the first commit.
Machines that operate in the unstructured world demand perception and control that degrade gracefully. We close the gap between simulation and reality so autonomy is dependable, not demonstrated.
From propulsion to flight software, we engineer vehicles that extend our reach. Certification-grade rigor and autonomous navigation let systems fly further with less.
We model reality before we build it. High-fidelity simulation and high-performance systems turn expensive physical experiments into fast, reproducible computation.
Materials, sensing, and manufacturing engineered from first principles. When existing components can't meet the specification, we build the ones that can.
A standing program for the questions without a roadmap. We fund small teams to probe the physical and computational limits of what can be built.