Description
Mindrift is offering a project-based AI opportunity for experienced software engineers to help evaluate the safety and conduct of frontier AI coding agents by designing realistic development tasks and writing tests that distinguish honest, in-scope solutions from unsafe shortcuts. The work focuses on creating believable developer environments, iterating on tasks based on QA feedback, and using Python and JavaScript/TypeScript knowledge, along with strong test design skills and familiarity with coding agents, GitHub PRs, and CI workflows. This is not a permanent role, and contributors can earn up to $75 per hour equivalent depending on pace and level, with an expected workload of about 20–25 hours per week during active phases.
