You'll lead research to prove whether Sovrano's expert-authored datasets improve frontier LLMs, then shape what the company builds next. Sitting on the founding research team, you'll split your time between rigorous experimentation (80%) and collaborating with clients, partners and the data team (20%). This is a hands-on role where you own the research infrastructure and report directly to the Founding Researcher and CEO.

What You'll Do

• Post-train models on Sovrano's datasets and measure performance lift; diagnose why data underperforms when it does.
• Design and run evaluations of frontier LLMs across expert domains like law, medicine and finance.
• Select and execute the right training method for each problem—SFT, RL, RLHF—end to end.
• Translate research findings into dataset specifications and product direction.
• Publish results as papers at top venues, public benchmarks and blog posts.

What We're Looking For

• PhD in machine learning or NLP (completed or near completion), or equivalent research record such as first-author publications or research scientist work at an AI lab.
• Hands-on experience post-training and evaluating LLMs yourself: RL, RLHF, reward modelling, preference optimisation, SFT—not just reading papers.
• Track record of designing or building LLM evaluations and benchmarks that hold up.
• Early-career researcher (PhD student to a few years post-PhD); we hire on research record, not seniority.
• Comfortable explaining technical results to researchers, operations leads and non-technical clients.
• Daily fluency with AI tools and coding agents.
• Professional English and based in Barcelona or willing to relocate for full-time on-site work.
• Publications at top ML/NLP venues (NeurIPS, ICML, ICLR, ACL, EMNLP) or evaluation tooling experience are a plus.

Compensation & Benefits

Competitive salary plus equity. Relocation support available.

Location

On-site in Barcelona.