Data Specialist

Be the careful eye behind our AI safety research.

Prague (on-site) • Part-time

We're a deep-tech AI safety stealth startup. We make powerful AI systems safe to deploy, a problem that gets more urgent as AI becomes more capable and autonomous. Backed by leading investors, with IOI/IMO medalists, top-university PhDs, best-paper winners, and advisors holding top prizes in CS and AI, including a Turing Award holder.

You'll look closely at what AI agents actually do: reviewing their transcripts, judging their behaviour, and building the datasets our research and customer work rely on. It's focused, detail-heavy work, and it suits someone who enjoys getting things exactly right.

What you'll work on

• Review transcripts of AI agents using tools and judge whether they behaved correctly, safely and as intended

• Build “gold” datasets: carefully labelled reference examples we use to evaluate AI systems and our automated judges

• Pre-select and prepare examples for customer reports, with a senior researcher giving final sign-off

• Spot patterns, edge cases and inconsistencies, flag them to researchers and help sharpen the guidelines

• Not a rubric-ticking labelling job: you'll learn what we're working on and use your own judgement when a case doesn't fit the rules

What we look for

• Reliable and high-integrity — you care about the quality of what you hand over; we'll trust your judgement

• Quick to understand — picks up new topics fast, notices nuance and small details

• Careful and patient — comfortable spending most of the week reading data closely

• Technical background — ideally you've completed at least one computer science or data science course

• Strong English reading comprehension

Bonus

• Some previous work experience, ideally where precision matters (banking, for example)

• Interest in AI and how modern AI systems are evaluated

• Able to start soon

What we offer

• A close look at how AI safety research works, side by side with a top-tier team

• Competitive compensation

• Part-time: 30 hours a week as a baseline, some weeks noticeably more. We guarantee an average of at least 30 hours a week every month

• Office in the centre of Prague

• Catered lunch and a fully stocked office kitchen

• Plenty to learn along the way, from how AI agents work to how they're evaluated