In a modern hospital setting overwhelmed with data, an innovative system for early sepsis detection is proposed, utilizing medical-grade wearable sensors and real-time data processing to monitor patient vitals and identify risks before a crisis occurs. This system employs an end-to-end inference pipeline involving Kafka for data streaming, DeltaStream for real-time feature engineering and inference, and a Generative AI model for clinical assessment, ultimately providing actionable alerts through various channels. The pipeline simulates real-world conditions using synthetic data, processes continuous vital sign streams to extract clinically relevant features, and utilizes a Large Language Model (LLM) to assess sepsis risk, delivering results with minimal latency for immediate clinical action. This streamlined approach eliminates the need for intermediate services, enhancing the architecture's efficiency and scalability, while the full implementation code is accessible for further exploration and validation.