Sepsis is a deadly condition where human body’s reaction to an infection causes organ damage and inflammation throughout the body. Quick detection of sepsis is vital because early detection of sepsis helps reduce mortality risk and makes treatment easier. Existing machine learning solutions for sepsis prediction are designed mostly for ICUs and employ non disease specific prediction models. In this work, we present a Machine Learning framework for early prediction of sepsis for non ICU patients. Individual Random Forests are trained for each of the four infection types – Pneumonia, Urinary Tract Infection UTI , Intra Abdominal Infection, and Skin and Soft Tissue Infection SSTI . Prediction is done based on structured information obtained from patients’ symptoms, vital sign measurements, and lab test results. Trained models are deployed as a Flask based web application which facilitates features like authentication of users, selection of disease type, prediction generation, confidence scoring, risk stratification and history management of predictions. Our system serves as an example for the potential application of ML driven systems for clinical decision support and demonstrates benefits of disease specific prediction approaches for early detection of sepsis. Prediction of Sepsis, Machine Learning, Random Forest, Healthcare Analytics, Flask, Disease specific Prediction, Clinical Decision Support. 43-47 Issue-4 Volume-10 Er. Yasmeen Viqar | Furqaan | Mehak Ayoub | Toiba Tahir