The objective of this work is to identify the hand written digits represented by rectangular box of 16x16 in a gray scale of 256 values. Backpropagation neural network (BPN) is one of the simplest models of supervised training multi layer neural networks. In this paper we design an BPN and train it with a set of hand written data. We also implement BPN on Master – Slave architecture to minimize the learning time. The performance parameters are evaluated for both sequential implementation and parallel implementation on Master – Slave architecture.