Leukemia patients who receive chemotherapy experience a decline in neutrophils and an increased risk of infections. Neutropenic sepsis is a life-threatening condition and a major cause of cancer-related mortality. Patients with neutropenic sepsis are generally treated with Broad Spectrum Antibiotics (BSA) as a first-line medication that destroys common causative organisms but may either miss the true pathogen or be overly broad leading to an increased risk of development of Antimicrobial Resistance (AMR). Physicians resort to using BSA due to a typical delay of 2-5 days for specific organism identification by blood cultures. We report the development and validation of an explainable AI powered system to predict bacterial growth in blood cultures (N=110) using readily available hematological parameters, enabling predictions 2-5 days ahead of actual culture results. Our best performing models yielded an accuracy and F1 score of 78%. In predicting gram-negative bacteria (GNB), the models demonstrated an accuracy and F1 score of 63%. To our knowledge, this is the first study to explore AI-powered early prediction of bacteremia in neutropenic sepsis patients in a South Asian population.