CareCoordEnv – Healthcare Care Coordination AI Agent
An OpenEnv environment for training and evaluating healthcare AI agents.
CareCoordEnv is an AI-powered healthcare care coordination environment designed to train and evaluate agents on realistic healthcare administrative workflows. It supports referral routing, prior authorization recovery, and post-discharge follow-up through structured actions, observations, and reward signals. The environment provides graded tasks for evaluating agent performance.
Technology stack
Problem
Healthcare coordination involves multiple connected workflows, including patient referrals, insurance authorization, and post-discharge follow-ups. AI agents need a structured environment where they can learn and demonstrate their ability to complete these multi-step tasks.
Approach
Designed an OpenEnv-compatible environment using typed Action, Observation, and Reward structures. Developed three tasks with different difficulty levels and task-specific graders. Integrated the environment with a FastAPI-based interface and prepared it for containerized deployment.
Solution
Developed CareCoordEnv with environment reset, step, and state operations. Implemented easy, medium, and hard healthcare coordination scenarios and an agent evaluation workflow. Packaged the environment for deployment through Hugging Face Spaces.
Results
Implemented three graded healthcare coordination tasks and an environment suitable for agent interaction and evaluation. Published the environment through a Hugging Face Space. Task completion rates and reward scores can be added from actual evaluation logs.