Freelance Physicist - AI Trainer
Mindrift • United States
Education
PhD
Type
Hourly
Pay
$55/hr equivalent
Mindrift pays per task and shows the pay as an hourly rate.
Last checked
5d ago
Checked
About this role
From the Mindrift listing
Please submit your resume in English and indicate your level of English proficiency. At Mindrift , innovation meets opportunity. We believe in using the power of collective intelligence to ethically shape the future of AI. What we do The Mindrift platform connects specialists with AI projects from major tech innovators. Our mission is to unlock the potential of Generative AI by tapping into real-world expertise from across the globe. Who we’re looking for: We’re looking for curious and intellectually proactive contributors, the kind of person who double-checks assumptions and plays devil’s advocate. Are you comfortable with ambiguity and complexity? Does an async, remote, flexible opportunity sound exciting? Would you like to learn how modern AI systems are tested and evaluated? This is a flexible, project-based opportunity well-suited for: Analysts, researchers, experienced professionals and consultants with strong critical thinking skills. People open to a part-time and non-permanent opportunity. About the project: You will create complex, realistic tasks that push frontier AI agents to their limits. Think scattered data, conditional procedures, and genuine domain expertise required. You'll build a detailed version with objective scoring, then write an ambiguous version intended to train the agent to succeed with less hand-holding. Real expert complexity only. You're improving the AI tools you'll eventually use yourself. If you have the relevant experience and are ready to take on this challenging and engaging project, join us! How to get started: Apply to this post and get the chance to contribute to projects aligned with your skills, on your own schedule. To begin working in production, you’ll need to complete the qualification step and project onboarding, where you’ll get familiar with the guidelines interface and try your first real task with guidance from our quality team. From creating training prompts to refining model responses, you'd be directly shaping how useful these models become for your own future work.
Requirements
- Must be eligible to work in United States
- Fluent proficiency in English (Written & Verbal)
- Reliable high-speed internet connection
- PhD's degree or equivalent professional experience
- Demonstrated expertise in STEM
AI Trainer interview prep
Questions AI Trainer interviewers often ask. They aren't taken from this listing.
How do you evaluate whether a model's response to a prompt satisfies the instruction, versus just sounding plausible?
I go back to the instruction line by line and check each requirement against the response individually, rather than reading the response holistically and asking if it feels right. A confident, well-written response can still quietly skip a constraint the prompt asked for.
What's your process for writing a rating rubric that two different reviewers would apply the same way?
I define each rating level with a concrete example rather than an abstract description, since two reviewers can read the same adjective differently but usually agree on whether a response matches a given example. I test the rubric on a handful of edge cases with another reviewer before rolling it out.
How do you decide when a prompt itself is ambiguous versus when the model's response is simply wrong?
I ask whether a reasonable person could interpret the prompt more than one way and still be following the instruction correctly. If the response fails under every reasonable reading, it's a model error; if it only fails under one specific reading, the prompt needs to be flagged as ambiguous instead.
How do you document the reasoning behind a rating so someone reviewing your work later understands the decision?
I write a short note pointing to the specific part of the response that drove the rating, so a reviewer can see what I was reacting to. A rating with no reasoning attached is hard to audit and even harder to disagree with productively.
Why this role
This Freelance Physicist - AI Trainer role pays $55/hr equivalent for AI training work that uses your STEM background. Onboarding covers the steps that are new to you.
Talent pool
We're light on STEM candidates
We've tracked 1,024 STEM listings and have few candidates to put forward for them. Set up a profile and we'll consider you for roles like this one.
Set up your profileSkills and categories
More jobs in these fields:
Related jobs
Verified reviews
Community reviews
Leave your review
Common questions
What does a typical Mindrift task involve?
Read a scenario, write a short prompt for the AI (around 100 words), evaluate two AI responses to it, fact-check the outputs, then write a brief explanation (around 50 words) on which response is better against the project rubric. The focus is clarity, safety, and rule-following, not creative writing.
Who is Mindrift good for?
Freelance writers, editors and generalist AI tutors with strong English and good research skills. You don't need a tech degree. Mindrift is built by the data company Toloka, and it also opens better-paid projects in fields such as cybersecurity, medicine and law to verified experts.
Is research-level AI training the same as data labeling?
It's closer to academic research. You write or check proofs, solve advanced problems and test the logic behind a model's step-by-step reasoning in your field.
Do I need a PhD for research-level AI training roles?
The best-paid roles usually expect a PhD or current enrollment. The domain assessment decides, though, and if you can solve the problems the degree matters less.
What does STEM work look like for a Freelance Physicist - AI Trainer?
You check AI answers to STEM questions, correct their errors and write expert reasoning. As a Freelance Physicist - AI Trainer, you're hired for that judgment, so hands-on STEM experience matters.
How much does this specific role pay?
This listing is posted at $55/hr equivalent, an hourly rate. Pay can change after we check the listing, so confirm the rate on the platform's application page.
Is a PhD required?
For this role, yes, or similar professional depth, and the assessment checks it. PhD candidates and people with published research have qualified without the degree.
Can I apply from outside the United States?
This role is open only to people based in the United States. The limit is usually about where the platform can legally hire, so applying from elsewhere rarely leads to an offer. Check the description for tax or right-to-work rules before you apply.