Freelance Coding Task Designer

Parsewave builds coding datasets for AI research labs. We're looking for experienced developers to design and validate complex, terminal-style coding tasks used in training and evaluating AI coding models.

Each task includes a problem description, a working solution (bash script), and automated test validation (pytest). You own the problem end-to-end, from writing the prompt to making sure the tests actually catch wrong answers.

Tasks need to be hard enough that current AI agents fail them. If a senior dev finds it easy, it's probably not good enough.

Good tasks look like real work: debugging a broken pipeline, implementing something non trivial from a spec, handling edge cases that matter in production. The goal is to find the gaps in what models can do and write problems that expose them clearly.

Fully remote, worldwide. No meetings, no timesheets, no minimum hours. 

Pick up tasks when it suits you. $80 per accepted task, $90 for excellent ones. 

Average task takes about 2 hours.