Job Description

Our team builds synthetic data and simulation tools & infra to train AI agents that use computers to complete everyday tasks. We are looking for a PhD intern to research how to generate tasks that require stronger planning and reasoning, with an initial focus on planning and navigation workflows.

The intern will design generation methods, build prototypes, and run experiments to understand which kinds of task difficulty help agents learn. The project combines research with practical engineering and offers an opportunity to contribute to training data for frontier models.

Responsibilities

Research methods for generating diverse, challenging reinforcement learning (RL) tasks with reliable success checks.

• Develop ways to measure and control task difficulty.
• Evaluate agent performance and distinguish reasoning failures from problems in tasks or environments.
• Build reproducible experiments and communicate findings and limitations.

Qualifications

Currently has, or is obtaining, a PhD in computer science or a related field

• Proficiency in Python
• Experience designing and implementing machine learning experiments
• Must obtain work authorization in the country of employment at the time of hire and maintain ongoing work authorization during employment Proficiency in HTML, TypeScript, Docker, SQL & AWS
• Experience with language models, reinforcement learning, AI agents, automated planning, synthetic data, or evaluation methods
• Intent to return to the degree program after the completion of the internship