Description
Mindrift is seeking experienced software developers to join a project-based initiative aimed at evaluating the safety and ethical behavior of AI coding agents. The primary focus is on designing challenging tasks that assess how agents handle complex scenarios, ensuring they adhere to guidelines rather than simply completing tasks. Responsibilities include creating realistic developer environments, crafting tasks with enticing shortcuts, developing comprehensive tests, and iterating based on feedback. This role emphasizes understanding proper code behavior and refining evaluations to ensure fairness and robustness.
