AI research lab
We specialize in autonomous behavior
We study how systems act on their own, check their own work, and keep a human in control. Rigorous work, measured before it is claimed.
What we study
- Multi-agent coordination
- Several agents share a goal and reach agreement without a central controller, so work keeps moving even when no one is in charge of everything. We study how to sustain that coordination when each agent knows only part of the problem and conditions change.
- Self-improving systems
- Systems that review their own output and correct course before they act, so quality improves without a person checking every step. We focus on when a system should trust its own correction and when it should pause and defer.
- Grounded decision-making
- Decisions anchored in evidence, with a clear trace of why they were made, so a person can review and trust the outcome. We work to keep that trace legible as the decision grows more complex.
A solid research foundation
Several experiments in progress, some of them confidential. The public research is the evidence that supports the work; what we do not show yet we keep private by design.
See the portfolio