Not just your agents: Reins AI understands that today's organizations need to adapt their processes, their systems, and their AIs to meet user needs. Our mission is to take current state and make it better, turning state-of-the-art methodologies into scalable practices for today's most critical use cases.
AI systems don't just reflect code. They reflect the people, structures, and decisions that shape them. We treat evaluation not as fault-finding, but as meaning-making. By making model behavior visible, we can create better organizations.
There is no IP or expertise blocking here. We build state-of-the-art evaluators so we can teach people the best of what we know.
Well-designed systems will have sufficient friction signals from use for us to make sure the system is in control, and notify you when it's not. And if the data isn't there, we partner with organizations that can create it easily.
Your data, your simulations, your agent interactions—they’re yours, always. No lock-in, no hidden harvesting, no secondhand monetization.
After years in AI, R&D, and consulting, I saw the same challenge repeated across industries: AI systems being deployed without truly understanding how they would behave in complex, real-world conditions with boundedly rational humans.
I started Reins AI to change that: to help organizations understand how their system is performing, or will perform, in real production environments. This isn’t about dashboards or demos: it’s about deeper understanding, stronger organizations, and more cooperative AI.
I started out as a computational linguist and spent over a decade building speech and language products at Nuance and Microsoft, from speech recognition in 45 languages to agentic systems for business applications. What stayed with me is how much of quality is decided after the model is built: whether a problem gets noticed, whether it reaches someone who can fix it, and whether anyone can show that the fix held.
At Reins AI, that’s my job: working out which signals from an agentic system deserve a person’s attention, getting them to the right person with the evidence attached, and turning what we find into fixes the build team can use. I’m based in Switzerland and work in English and German.
I've worked in various tech and consulting roles, but I started out as a military weather forecaster, leading both operational and data/modeling teams supporting live missions. I learned that decision makers still need to act when the model is uncertain, and doing that well takes an honest accounting of how the forecast system performs.
At Reins, that's my job: measuring how AI systems really behave, and building the architecture that supports better decisions.
Coming soon.
