LLM DATA RESEARCHER – Sovrano AI (Barcelona, full time)
Prove, with evidence, what expert data does to a model, and decide what Sovrano builds next.
ABOUT SOVRANO
Sovrano AI is a data and research lab in Barcelona. We bring real experts into building AI: lawyers, doctors, engineers and tax advisors do the work, we capture their judgment and turn it into the post-training data, rubrics and benchmarks that frontier AI labs, robotics companies and enterprise AI teams use to train and evaluate their models. Behind it: a network of 400,000+ professionals, 100+ partner universities across Europe, 23 languages and EuroExec, our own benchmark.
ABOUT THE ROLE
This role answers, with evidence, whether our data makes models better, and then tells the team what to do about it. The week is about 80% research (training, evaluating, publishing) and 20% communication and coordination with clients, marketing and the data team. You report to Pau Arnal, Founding Researcher, and work directly with him and the CEO. This is a founding seat on the research team: an agenda that is genuinely yours to shape, and results that change what the company does the following week.
WHAT YOU WILL DO
- Prove the data works: post-train models on our datasets and measure the lift. When the data is not pulling its weight, find out why.
- Build the benchmarks: design and run evaluations of frontier LLMs on expert domains.
- Train the models:



pick the method each question needs and run it end to end.
- Shape the next dataset spec with the data team, and verify the quality of what ships.
- Explain the evidence to researchers, operations leads and clients.
- Publish: papers at top venues, public benchmarks, blog posts, with your name on the work.
WHAT WE OFFER
Competitive salary plus equity. Full time, permanent, on-site in Barcelona. Help with relocation if you are moving here.

Requisits

- PhD (or close to finishing), or a research record that does the same job, in machine learning or NLP.
- Hands-on experience post-training and evaluating LLMs yourself, beyond reading the papers. You know where evaluations go wrong.
- You move fast: you run the experiment early and write up what you found, even when it is the opposite of what you hoped.
- You can explain results to a researcher, an operations lead and a client.
- AI native: you work with AI tools and coding agents daily.
- Professional English.
- Based in Barcelona or willing to relocate (on-site role).
Nice to have: publications at top venues; experience building evaluation tooling; a second European language or interest in a professional domain (law, medicine, finance); a strong opinion about which benchmark is overrated.
To apply: CV, a link to the research work you are proudest of, and one sentence on a benchmark you think is overrated, and why. No cover letter.

📌 llm data researcher – sovrano ai (Barcelona)
🏢 Sovrano AI
📍 Barcelona