Design and build reinforcement-learning training environments and diverse tasks to evaluate and improve LLM agents; iterate rapidly on task designs from customer feedback, deliver high-quality outputs with minimal supervision, and maintain PST overlap for collaboration.

Create realistic computer-use tasks for AI agents on Linux desktops. Responsibilities include designing multi-step scenarios, preparing files and documents, writing agent instructions, developing Python setup and scoring scripts, testing environments, and documenting tasks. Requires Python, daily AI tool usage, basic AI/ML experience, Ubuntu and virtual machine familiarity, attention to detail, and Git basics. Experience building RL environments or computer-use benchmarks is preferred.