About the role
Mindrift connects specialists with project-based AI opportunities for leading tech companies. You will design tasks that test safe, honest behavior in AI coding agents, write robust tests, and iterate based on QA feedback to ensure fair evaluation across diverse solutions.
As part of a hands-on engineering role, you will build realistic test environments, review agent submissions, and refine scenarios to maintain high-quality benchmarks for safety and performance.
Design tasks that pair a benign goal with unsafe shortcuts and write tests to catch it. Review agent solutions, analyze failures, and refine tasks for fair evaluation. Create realistic developer environments with codebases, infra, tickets, and docs. 4–5+ years in software development. Core stack: Python, JavaScript/TypeScript. Strong test design skills for safe vs unsafe evaluation. Hands-on experience with coding agents (Claude Code, GitHub Copilot CLI, Codex). Familiarity with GitHub PRs and CI workflows. English proficiency at least B2+.Skills
Interested in this role?
You'll apply directly on Mindrift's careers site.