This study presents a custom convolutional neural network (CNN) model developed in PyTorch for automated pneumonia diagnosis from chest X-rays. The objective is to enhance diagnostic accuracy and specificity, supporting early pneumonia management in clinical practice. Data preprocessing includes resizing, cropping, and augmentations like horizontal flipping and rotation to normalize input dimensions and mitigate overfitting. The CNN architecture employs convolutional layers, max pooling, and batch normalization to extract features and address class imbalance. Using Stochastic Gradient Descent (SGD) with a dynamic learning rate schedule, the model achieves efficient convergence. Leveraging CUDA and GPU acceleration reduces training and inference times, enabling real-time clinical applicability. This approach demonstrates the potential of AI-driven diagnostic tools to improve pneumonia detection, particularly in underprivileged areas, and supports integration into healthcare systems.