Blood cancers, such as leukemia, lymphoma, and myeloma, continue to pose a significant global health concern, accounting for roughly 6% of all cancer cases globally (WHO, 2021). Despite advances in medical treatment, early identification remains critical in increasing survival rates. Current diagnostic methods, such as microscopic study of blood smears, are susceptible to human error and inefficiency. Convolutional Neural Networks (CNNs) have transformed medical image analysis by delivering precise and automated methods for disease identification. This paper proposes a CNN-based model for identifying specifically Acute Lymphoblastic Leukemia (ALL) from large datasets of blood smear pictures. To improve accuracy, our model combines image preprocessing, feature extraction using transfer learning, and ensemble methods. We also address major issues in medical imaging, such as data inconsistency and limited datasets, using data harmonization and transfer learning. Evaluation criteria such as accuracy, precision, and recall reveal the model’s strong performance, making it a potential approach for the early identification of blood cancers in clinical settings.