Scalable and Fault-Tolerant Container-based Architecture for Early Sepsis Detection through Integrated AI

Sepsis is a critical medical condition characterized by an overreaction of the immune system to infections, leading to potentially fatal outcomes if not promptly treated. This study presents the development of an AI-based application designed to predict sepsis risk using electronic medical records. The original monolithic architecture of the application, while functional, lacked the necessary scalability, fault tolerance, and availability required for modern healthcare applications. To address these limitations, a container-based architecture was proposed and implemented using Docker Swarm, leveraging its inherent scalability and fault-tolerant capabilities. Comparative analysis with Kubernetes is planned to determine the optimal orchestration tool for this application. Preliminary results indicate that LightGBM outperforms other models in prediction accuracy. Future work includes extensive testing with synthetic data and stress testing tools to evaluate the proposed architecture’s performance under various conditions. This research aims to enhance early sepsis detection and improve patient outcomes by providing a robust, scalable, and reliable AI-driven solution.