Physics AI Training Expert

About OpenTrain

OpenTrain is the #1 platform for building careers in AI training and data labeling. We connect contributors with focused, short-term projects that let you grow an AI training portfolio, work remotely, and shape how modern AI systems learn from people.

For this role, OpenTrain is hiring and contracting directly — you'll join a distributed team working on cutting-edge evaluation tasks that matter to researchers and engineers building large language models.

Why AI training matters

AI training (a.k.a. data labeling or human feedback) is the human work that teaches models to reason and respond. Contributors write examples, rate outputs, and create evaluation benchmarks — the results directly influence how state-of-the-art systems behave.

This role places you at the intersection of physics and model evaluation: your work will help define benchmarks that test model understanding across undergraduate to PhD-level topics.

The role

As a Physics AI Training Expert you will design challenging physics problems, produce high-quality, step-by-step solutions, and contribute structured annotations and feedback used to evaluate and improve large language models.

You will collaborate with other evaluators and researchers to align prompts and problems with evaluation goals and help refine new physics benchmark sets spanning early undergraduate through PhD-level topics.

What you'll do

• Design physics problems that probe LLM reasoning limits across mechanics, electromagnetism, thermodynamics, quantum mechanics, and related areas.
• Write clear, logically organized step-by-step solutions that can serve as ground truth and model training examples.
• Create detailed annotations, error analyses, and constructive feedback to support evaluation and grader consistency.
• Align problem prompts and solution formats with specified evaluation goals and benchmark criteria.
• Help define and expand benchmark sets for multiple academic levels (undergraduate to PhD).

Requirements

You must demonstrate advanced physics knowledge and the ability to solve multi-step problems with transparent, rigorous reasoning.

• Advanced physics background (strong undergraduate or graduate-level study; engineering entrance-exam level or higher).
• Proven ability to explain complex physics concepts in simple, structured language and to write step-by-step technical explanations.
• Strong Python skills for scientific computing (numerical solutions, basic scripts, or notebooks).
• Analytical and research skills, attention to detail, and experience with structured remote collaboration.
• Fluent English comprehension for reading, writing, and reviewing technical content.

Work setup & schedule

This is a fully remote contractor assignment. The engagement runs for 3 weeks and requires at least 20 hours per week with a minimum of 4 hours of work per day.

You must overlap 4 hours with Pacific Standard Time (PST) each day for synchronous collaboration and reviews. OpenTrain contributors manage their own time within those constraints.

Who should apply

Apply if you enjoy writing clear technical solutions, want to shape how LLMs understand physics, and can meet the time and collaboration requirements. This role fits advanced students, researchers, or practitioners comfortable with physics problem-solving and Python.

• Ideal for physics graduate students, instructors, teaching assistants, researchers, or engineers with strong explanatory skills.
• Good fit for people seeking focused, short-term contractor work that contributes directly to AI evaluation research.

How the process works

Create an OpenTrain account (free) to apply and build a profile showcasing relevant physics and Python experience. If selected, you'll be contracted by OpenTrain and provided task guidelines, evaluation rubrics, and communication channels for collaboration.

OpenTrain handles project onboarding, task distribution, and payment according to the contract; you deliver problems, solutions, and annotations per the provided schedule and quality standards.