Mechanize, Inc.
Mechanize builds reinforcement-learning environments that AI labs use to train and evaluate code-generation models, and licenses these environments as products and services. It primarily serves frontier AI research groups and model teams, providing tasks, benchmarks, and infrastructure tailored for coding agents. The privately held company is headquartered in San Francisco and has been covered by the New York Times; backers include Patrick Collison, Nat Friedman, Daniel Gross, Jeff Dean, and others.
Open Positions at Mechanize, Inc.
4 open positions
Build sophisticated reinforcement-learning environments that expose and help fix the complex failures of coding models. You’ll own tasks end to end while directing coding agents, analyzing model behavior, and contributing to shared engineering infrastructure.
Design and build reinforcement learning environments, coding challenges, and automated graders that reveal meaningful gaps in frontier coding models. This independent, high-ownership role combines software engineering, task evaluation, and careful analysis of coding-agent behavior.
Design and build reinforcement learning tasks and automated grading environments that expose weaknesses in frontier coding models. This high-ownership internship combines Python software engineering, creative problem-solving, and careful analysis of AI-agent behavior.
Build and refine reinforcement learning environments and automated graders that expose weaknesses in frontier coding models. This independent, high-ownership role is designed for capable recent graduates who can code in Python and think creatively about model behavior.