Software Engineering Intern

Responsibilities

• Design, build, and refine reinforcement learning (RL) tasks used to train and evaluate frontier coding models
• Own the full lifecycle of a task, including ideation, implementing grading infrastructure, and running models against tasks
• Analyze model failures to iterate on task rigor and fairness
• Use coding agents heavily to develop tasks and evaluate their outputs

Requirements

• Graduating in 2026 or 2027
• Proficiency in Python
• Ability to work independently in a high-ownership environment
• Strong technical fundamentals and intuition for AI model behavior

Preferred Qualifications

• Significant experience working with frontier coding agents
• Ability to identify subtle gaps in model capabilities and anticipate where models might take shortcuts

Benefits

• Competitive compensation at $100 per hour
• 401k
• Health, dental, vision, and life insurance

About the Company

Mechanize is a ~20 person team based in San Francisco that builds reinforcement learning environments for frontier AI labs. We focus on creating challenges that expose the complex, judgment-heavy gaps in current coding models. We are backed by prominent industry leaders including Patrick Collison, Nat Friedman, and Jeff Dean.