This python project uses neural network of several layers to train and predict the MNIST dataset of handwritten digit images. Neural network is created with the help of keras and tensorflow and takes few minutes to train. Once trained the NN-model is automatically saved and the test data can be predicted using saved model.