Physics AI Trainer

When AI gets the details wrong in your field, real people pay for it. The people who know the work best have to be the ones training these models. You’re not being replaced by AI. You’re making sure it’s used accurately and safely.

• Design realistic, challenging physics and engineering tasks based on professional workflows, including scenarios, colleague-style prompts, and supporting files such as test data, specifications, simulation outputs, lab records, schematics, drawings, and reports.
• Run tasks through frontier AI models and evaluate calculations, analyses, design reviews, and technical reports against professional standards.
• Compare model outputs generated from identical prompts and files, determine which performed better, and document strengths and shortcomings.
• Write detailed grading rubrics defining the assumptions, methods, magnitudes, units, failure modes, and deliverables required for a correct response.
• Identify and document concrete model failures, including unit or scaling errors, misread data, unsupported conclusions, fabricated or ignored source material, missed safety or boundary conditions, and off-brief interpretations.
• Build professional-style task files yourself, ranging from compact multi-file scenarios to larger assignments that may take several days to develop.
• Review, refine, and contribute to tasks created by other subject-matter experts across physics and adjacent disciplines.
• Complete a proposed task based on professional experience, including the scenario, prompt, source files, anticipated model failure points, and full task build after proposal approval.

• 2+ years of applied experience preferred in physics, mathematics, chemistry, biology, engineering, data science, or a closely related technical discipline.
• In-progress Bachelor's degree or higher; 2+ years of experience outside undergraduate study is preferred.
• Current or recent hands-on individual-contributor experience in a technical field, rather than solely managerial experience.
• Experience working with real technical data and field-specific tools, such as instrument or test data, measurement files, simulation outputs, schematics, drawings, and analysis software.
• Strong scientific and experimental reasoning skills, including experiment design, measurement error, uncertainty, and common statistical pitfalls.
• Ability to assess work in adjacent sub-disciplines and explain what was done correctly or incorrectly.
• Strong written English communication skills and the ability to clearly explain why a technical result is incorrect.
• General familiarity with AI and LLM tools and the ability to distinguish well-reasoned outputs from plausible but incorrect answers.
• Coding or scientific computing experience in Python or a similar language is strongly preferred and required for some projects; comfort with command-line tools is also needed for coding-based work.
• Baseline technical proficiency with cloud file tools, browser profiles, desktop application installation, and common file-management tasks.
• Advanced degree or professional licensure is a plus but not required; practical applied experience is valued over credentials.
• Must be able to complete the skills assessment independently without using AI tools.

• Pay of $40-$125 per hour USD, paid via PayPal on a regular cadence.
• Fully remote contract work with flexible scheduling.
• Minimum commitment of 10 hours per week, with no weekly maximum; many projects prioritize 30-40 hours per week and full-time hours may be available.
• Multi-day task timers that allow work to be spread across several days.
• Project access to Claude and ChatGPT; no personal subscription is required.
• Structured onboarding, training, expert review, written feedback, and opportunities to review other experts' work.
• Compensation for each completed stage of the project process, subject to acceptance requirements.

DataAnnotation offers remote contract roles matched to your skills. Experts help AI perform better by reviewing responses, checking accuracy, refining prompts, and rating outputs in their own field. Consistent projects, flexible hours, and expert rates, with work judged on output quality rather than hours logged. Get paid to teach AI your expertise.