A RL project in healthcare: adjusts sepsis treatment dosages (e.g., fluids, vasopressors) based on patient's blood pressure and lactate levels. Learns from simulated ICU data to maximize survival rates, using rewards for stable vitals and penalties for organ failure.
🎯 Motivation India's ER Crisis: 2.2M sepsis deaths annually, 70% from triage delays in public hospitals. Current systems fail because:
Resource scarcity → Limited ICU beds, staff shortages Dynamic priorities → Patient conditions change rapidly Bias risks → Age/gender disparities in treatment SepsisCare-Auditor tests if LLMs can make life-critical decisions under real Indian hospital constraints.
📚 Data Source MIMIC-III inspired synthetic dataset simulating Delhi public hospital ERs:
Built With
- docker
- face
- fastapi
- gymnasium
- hugging
- openenv
Log in or sign up for Devpost to join the conversation.