Facial age and gender recognition has been relevant to a lot of applications, particularly since the social networking platform became popular. Since recent advances in convolutional neural network (CNN), a significant increase in performance can be obtained on this task. In this project, I will explore how to build a CNN for age and gender classification as well as evaluate the performance of the model. The data used to train and evaluate the model are real-world images which is challenging but can better reflect real-world problems. A demo application is also developed which support age and gender detection in real-time. The purpose of this report is to demonstrate the techniques that are used, the design decisions, and the analysis of the results obtained from experiments.