Developed and delivered a convolutional neural network model for hematologic disease analysis within the apache/singa repository, focusing on modularity and extensibility for rapid prototyping. The work involved designing a complete CNN architecture in Python, incorporating convolutional layers, batch normalization, pooling, and fully connected layers, all optimized with a softmax cross-entropy loss function. A factory function was implemented to streamline model instantiation, and a set of supported optimizers was defined to standardize training workflows. This contribution emphasized reusable deep learning components, enabling more efficient experimentation and evaluation of hematology image analysis tasks using core machine learning and neural network techniques.