Sepsis is a life-threatening condition that occurs when the body’s response to infection triggers organ dysfunction. It is a major global health problem, affecting millions of patients each year and contributing to high rates of morbidity and mortality in intensive care units. Early detection is critical, as every hour of delayed treatment significantly increases the risk of death. This project aims to develop a deep learning model to predict sepsis in ICU patients before clinical diagnosis. We engineer and compare three deep learning architectures: Gated Recurrent Unit with Decay (GRU-D), Temporal Convolutional Network (TCN), and Temporal Residual Transformer (TRT).