RESEARCH AND FRONTIER
Credentialed at the edge.
Separately from client work, we lead AI alignment research collaborations with frontier AI labs. That research keeps the systems we ship grounded in what is actually in production, not copied from last year's blog posts.
THE WORK
Applied AI alignment research
We lead small teams of data scientists and ML engineers on research collaborations focused on making frontier models more legible and safer to deploy. The work is applied: interventions that can be validated on real models, techniques that transfer to production systems.
We keep the specifics of our collaborators' roadmaps off the public record. What clients get is a practice grounded in the current shape of the field.
WHY THIS MATTERS FOR CLIENT WORK
The field moves every quarter; we move with it
A client stack designed six months ago against last year's best practices is a stack that has already started aging. The research work keeps our default patterns current.
It also means when a client asks whether a new capability is production-ready, or whether to wait a quarter, we can answer with something better than guesswork.
COLLABORATION
We are open to new research partnerships
If you are a research lab, independent researcher, or applied team working on alignment, interpretability, or evaluations and want a technical partner, reach out.
We do not publish a list of current collaborators. We do reply to serious inquiries.
Talk to us about research.
Tell us what you are trying to do. We reply within one business day.
Talk to us