Handwritten digit recognition is an important area in pattern recognition and computer vision, with widespread applications in bank cheque processing, postal services, form digitization, and automated data entry systems. One of the main challenges in this field is the large variation in individual handwriting styles, including differences in size, shape, and orientation of digits. Traditional handwritten digit recognition techniques rely heavily on manual feature extraction, which is time-consuming, requires expert knowledge, and often results in lower accuracy and limited adaptability to diverse handwriting patterns. To overcome these limitations, a web-based handwritten digit recognition system is implemented using a Convolutional Neural Network (CNN). The system is demonstrated through a simple website developed using a lightweight backend framework such as Flask, which allows users to upload handwritten digit images and view prediction results easily. The CNN model is trained and tested using the MNIST dataset, consisting of digit samples ranging from 0 to 9. By automatically learning important features through convolutional, pooling, and fully connected layers, the proposed system eliminates manual feature extraction and achieves improved accuracy and faster processing. The results confirm that the CNN-based approach is efficient, reliable, and suitable for real-time and practical handwritten digit recognition applications.