STEM Expert, AI Training Sprint (12 Weeks)

This is a 12-week sprint where you put your professional judgment into that work: you write, review, and correct model output in your own field, so the model learns how someone who actually does the job thinks rather than how a generalist guesses.

The work runs across mathematics, physics, chemistry, biology, and engineering analysis. Most of it turns on the working rather than the answer: setting up the problem correctly, spotting where a derivation goes wrong, and writing the correction so it teaches. Depth in one field matters more here than breadth across several.

You work fully remote as an independent contractor, 30 to 40 hours a week, on a schedule you set yourself. Tasks are claimed from the platform. There are no fixed shifts and no standing meetings. Hours and quality are what count.

You are not being replaced by AI. You are training it, so that when someone in your field relies on it, the answer holds up. Human experts have to be the ones producing this data. AI training AI does not give you something anyone should trust.

KEY RESPONSIBILITIES Work through tasks on our partner's platform, applying the same judgment you would apply on the job; Write and review model responses, explanations, and reference material to a professional standard; Correct model output where it is wrong and set out clearly why it is wrong; Produce context files and worked examples the model can learn from; Act on reviewer feedback and hold your quality across the full 12 weeks; Log 30 to 40 hours a week and keep your throughput steady; Work from your own expertise, without LLM assistance on projects that do not permit it. IDEAL QUALIFICATIONS PhD or Masters in a STEM field, completed or in progress; At least 3 years of non-academic professional experience applying it; Genuine depth in at least one of mathematics, physics, chemistry, biology, or engineering; Strong technical writing, including derivations and worked solutions; Based in the United States, United Kingdom, Canada, or Australia. NICE TO HAVE Peer-reviewed publications; Industry R&D experience alongside the academic background; Teaching or tutoring experience at university level; Prior work on an AI training or data platform. WHAT SUCCESS LOOKS LIKE You clear the Starter Assessment on your own merit; You hold 30 to 40 hours every week for the full 12 weeks, with no quiet drop-off in week six; Your written explanations are clear enough that a colleague in your field would sign off on them; You take reviewer feedback and apply it rather than argue it; Your context files are accurate, well sourced, and usable as training data; You finish the sprint, and then decide for yourself whether to keep going. CONTRACT & PAYMENT TERMS

$125 per hour for 30 to 40 hours a week, which works out at roughly $3,750 to $5,000 a week and $45,000 to $60,000 across the full 12-week sprint.

Independent contractor, fully remote. Payment runs through PayPal, so you need a PayPal account. Neither Sovrano AI or its partners file or submit tax forms on your behalf.

Getting on: a Starter Assessment that takes most people three to four hours. It pays a flat $240 if you pass. It is the only route onto the platform, so do the research and show your best thinking.

Quality: the bar holds for the whole sprint. If your work falls below it you are offboarded and paid for the hours you have already worked, with nothing guaranteed beyond that.

Time off: disclosed vacation is fine as long as you make the hours up inside the 12 weeks, within three weeks of the break. Flag planned time off as early as you can.

After the 12 weeks: your call entirely. Carry on with other projects at whatever cadence suits you, take only the ones that interest you, or stop. No obligation either way.