Sovrano sets the human standard for AI.

A machine should not decide what good work looks like. A person who does the work should. We build the RL environments, benchmarks and expert data that put that judgment into training.

How we got here.

2022

CareerOS founded in Barcelona

We started in careers, working with universities to help their students and graduates land their first roles.
2025

Reach across 200+ universities

Partnerships with 200+ universities in 25+ countries gave us reach into finance, law, engineering and medicine, the fields labs now need judgment from.
2026

Sovrano AI launches

We turned that reach into an expert bench. Practitioners now author and grade the data AI labs train on, and we published EuroExec, a 598-task benchmark double-judged by 47 experts.

Why labs work with us

Experts, verified.

Every expert is checked against their credentials and tested in their domain before any paid work.

Graded twice.

Benchmark work is double-judged and we publish the agreement. On EuroExec, 76% of rubric scores land within ±1 point between two experts.

Written where the work happens.

Tasks are authored by practitioners in the market they test. The right answer in Frankfurt is not the right answer in Madrid.

Human judgment stays in charge. Europe is how we prove it.

Law, markets and institutions shape how work gets done in each country. We recruit credentialed practitioners who work under those rules every day. Their judgment lives in practice, so it has to be authored by them.

Universities: 200+
Countries: 25+
Data residency: EU

The people behind Sovrano AI.

The team that built CareerOS. Its university partnerships gave us reach into every field a lab cares about. Today that reach is an expert bench: practitioners who author and grade the data AI companies train on, with students learning the work alongside them.