About
An AI research and product company
Horizon Mind is early. We are a research-driven company working on intelligent systems and on the security of those systems — the models, the agents, and the infrastructure they run on.
Our view is that these are not separate problems. A model that behaves well in evaluation and badly under adversarial input has not been finished. An agent without clear boundaries is not a product. We would rather build fewer things and understand them properly.
The path we work along is research, then prototype, then product. We publish what we can, and we say plainly where something is still an experiment.
Manifesto
Intelligent systems that can be trusted with real work
The vision is a future where AI systems are relied on for work that matters, and where that reliance is earned rather than assumed. That requires knowing how a system behaves when conditions are not favourable — under adversarial input, at the edges of its training, and when it is handed permissions it could misuse.
The mission is narrower and nearer: research those behaviours, and turn what we learn into systems other people can depend on. Security is part of the design rather than a review at the end, because the properties that matter most are the ones decided while the system is being built.
We expect the work to widen over time — AI engineering, computer vision, agent development and security consulting are all directions the research points at. It widens as the research supports it, not ahead of it.
Approach
How we work
- Research-driven
- We turn research into practical technology.
- Security-first
- Security is part of the system, not an afterthought.
- Engineering-focused
- We care about robust, scalable, production-ready systems.
- Product-minded
- We build technology that solves real problems.
- Long-term thinking
- We develop systems designed for an evolving AI landscape.
Team
Small, early, and building
Horizon Mind is at the stage where the team is small enough that everyone works across research and engineering rather than on one side of the line. That is deliberate for now: the questions we are interested in are not answered by a paper alone, and they are not answered by shipping alone either.
We will introduce the people here once there is a team page worth reading rather than a set of placeholder cards. In the meantime, the clearest picture of how we think is the research and the principles above.
Careers
Working here
We hire slowly and we post a role only once it is real and funded. If you work on adversarial robustness, agent security, privacy-preserving inference, interpretability, or the engineering that carries those into production, we would like to hear from you regardless of whether something is open.
Write to us through the contact page with what you have worked on and what you want to work on next. A link to a paper, a repository or a write-up is worth more to us than a résumé.
No open roles right now
Positions are listed here when they open, with the actual scope and level. We do not keep speculative listings up to collect applications.
Let's build what comes next.
If you are working on something where intelligence and security both matter, we would like to hear about it.