Create Complex STEM Problems for AI Evaluation
Budget: $250 – $750 USD
Project Title: Wanted: STEM Experts for 20 Original Advanced Problems (Robotics, CV, RL, STEM)
Project Description:
We are seeking highly skilled researchers, algorithm engineers, and mathematicians to author 20 highly complex, completely original STEM problems. These problems will be used to evaluate the reasoning capabilities of frontier AI models, so they need to be genuinely difficult.
The Deliverables (Format depends on your domain):
You will deliver 20 original problem tasks. The required format varies based on the subject:
For General STEM (Mathematics, Physics): Deliver a clean LaTeX file containing one original, self-contained problem, the uniquely determined correct final answer, and a step-by-step rigorous proof/worked solution.
For Computer Science Domains (Robotics/Control, RL, Computer Vision): Deliver both a LaTeX file containing the mathematical formulation/reference solution AND a single Python file with clean, bug-free, executable code that solves the problem.
Accepted Domains (Pick whatever you are strongest in):
General STEM (Advanced Mathematics, Physics, etc.)
Robotics and Control Systems
Reinforcement Learning
Computer Vision
Strict Quality Requirements (Non-Negotiable):
100% Originality: The problem, solution, and code must be created entirely by you.
No AI Assistance: You cannot use ChatGPT, Claude, or any other LLM to generate the prompt, code, or solution (this is strictly checked and is grounds for rejection).
No Plagiarism: You cannot copy problems from textbooks, research papers, LeetCode, or existing datasets.
Ideal Background:
Graduate degree (MS/PhD or in progress) in Computer Science, Mathematics, Engineering, or a closely related field; advanced research or competitive programming experience; comfort with LaTeX and Python; and the ability to write clear, rigorous solutions and algorithms.
Difficulty level of the problem will be university level master or PhD level.
How to Apply (Please Read Carefully):
To apply, please include the following in your proposal:
Your academic/research background and areas of specialization.
Your availability and how many problems per week you could produce.
For validity purposes, DO NOT attach a sample to your bid. If your profile is a match, we will initiate a chat with you. You will then be required to provide ONE (1) short sample problem (in LaTeX, or LaTeX + Python depending on your domain) directly in the chat at roughly graduate difficulty — this sample is the main thing we will evaluate to award the project.
Note: We will share a detailed brief (formatting rules, full quality checklist, and example problem profiles) with shortlisted candidates in the chat. Generic applications or bids that look AI-generated will be skipped.
Budget & Payment:
If fewer than 20 are delivered (or accepted after our quality/originality checks), payment will be prorated to the number of completed tasks. Payment will be released after the final verification and testing of the delivered questions.
Project Description:
We are seeking highly skilled researchers, algorithm engineers, and mathematicians to author 20 highly complex, completely original STEM problems. These problems will be used to evaluate the reasoning capabilities of frontier AI models, so they need to be genuinely difficult.
The Deliverables (Format depends on your domain):
You will deliver 20 original problem tasks. The required format varies based on the subject:
For General STEM (Mathematics, Physics): Deliver a clean LaTeX file containing one original, self-contained problem, the uniquely determined correct final answer, and a step-by-step rigorous proof/worked solution.
For Computer Science Domains (Robotics/Control, RL, Computer Vision): Deliver both a LaTeX file containing the mathematical formulation/reference solution AND a single Python file with clean, bug-free, executable code that solves the problem.
Accepted Domains (Pick whatever you are strongest in):
General STEM (Advanced Mathematics, Physics, etc.)
Robotics and Control Systems
Reinforcement Learning
Computer Vision
Strict Quality Requirements (Non-Negotiable):
100% Originality: The problem, solution, and code must be created entirely by you.
No AI Assistance: You cannot use ChatGPT, Claude, or any other LLM to generate the prompt, code, or solution (this is strictly checked and is grounds for rejection).
No Plagiarism: You cannot copy problems from textbooks, research papers, LeetCode, or existing datasets.
Ideal Background:
Graduate degree (MS/PhD or in progress) in Computer Science, Mathematics, Engineering, or a closely related field; advanced research or competitive programming experience; comfort with LaTeX and Python; and the ability to write clear, rigorous solutions and algorithms.
Difficulty level of the problem will be university level master or PhD level.
How to Apply (Please Read Carefully):
To apply, please include the following in your proposal:
Your academic/research background and areas of specialization.
Your availability and how many problems per week you could produce.
For validity purposes, DO NOT attach a sample to your bid. If your profile is a match, we will initiate a chat with you. You will then be required to provide ONE (1) short sample problem (in LaTeX, or LaTeX + Python depending on your domain) directly in the chat at roughly graduate difficulty — this sample is the main thing we will evaluate to award the project.
Note: We will share a detailed brief (formatting rules, full quality checklist, and example problem profiles) with shortlisted candidates in the chat. Generic applications or bids that look AI-generated will be skipped.
Budget & Payment:
If fewer than 20 are delivered (or accepted after our quality/originality checks), payment will be prorated to the number of completed tasks. Payment will be released after the final verification and testing of the delivered questions.