https://www.blotdle.com/
We're Blotdle. We build science benchmarks for testing AI models, and we pay researchers to turn a piece of their own computational work into a task an AI agent has to solve.
A task is one problem from your own work, such as a large PDE solve that has to meet an accuracy target or a mixed-integer program with a certified optimality gap. Formal proofs in Lean count too. This is not a comprehensive list; we are open to various types of work in your field. A program checks the answer. It has to be something current AI models still get wrong, and the answer can't already be published.
Requirements
• MS or PhD in applied mathematics, statistics, operations research or a related field, or currently in a PhD program
• 2+ years of research experience, in academia or industry
• You write your own code in Python, Julia, R, MATLAB or similar, or proofs in Lean 4
• Hands-on experience with numerical PDEs, integer or derivative-free optimization, survey or causal statistics, or Mathlib
• You can point to a step in your own work where a careful-looking method still gets the wrong answer, and explain why
• Clear technical writing in English
Preferred
• Peer-reviewed publications or a Google Scholar profile
• Comfortable on the Linux command line. Docker is a plus.
$75 to $400 per accepted task, paid only on acceptance. Remote contract, your own hours.
Pay: From $50.00 per hour
Work Location: Remote