This project is about a software that recognizes digits of Nepali language i.e. Devanagi digits. When given a set of Nepali digits from 0-9 it easily recognizes them. It can be further developed to recognize characters or words of Nepali languages. It implements theconcept of neural network that includes Back-propagation and Gradient descent Algorithm.In general perspective, one of the primary means by which computers are endowed with human like abilities is through the use of a neural network. Neural networks are particularly useful for solving problems that cannot be expressed as a series of steps, such as recognizing patterns, classifying them into groups, series prediction and data mining.Therefore, with an aim to make computer recognize the digits, this project have been developed.A neural network trained for classification has been designed to take input samples and classify them into groups of digits. The accuracy obtained for classifying handwritten Devanagari digits is 96.7%.At first, data (images) that are needed for the training and testing are collected andpreprocessing is done. Then after, those training data set are supplied as a means of input and these are computed through different nodes of the neural network to produce certain outputs. Obtained output is compared with the target value, and error is calculated. The implementation of back propagation comes at the time of back-propagating error produced and gradient descent is used to minimize the error to update the link weight accordingly. The training process continues until the system produce the output that is same as the target value. Testing process is simply done by providing pixels values of image as an input to the input nodes that need to be predicated, and the output is compared with the target value and result about predication is displayed. Testing process does not involve the back-propagation and gradient descent.