Lung diseases like COVID-19, Pneumonia, and Tuberculosis pose a serious health risk to the global human population. These conditions lead to huge loss of human life as well as economic resources on an annual basis. Therefore, there is the need to detect and treat these diseases at an early stage. To reduce the ever-increasing burden on the limited healthcare resources, this study proposes an automated technique to predict Pneumonia, Tuberculosis, and COVID-19 with chest X-ray radiographs. The proposed technique uses a customized deep learning model trained on over 16,000 chest X-ray images collected from various publicly available datasets. Data augmentation methods have been applied before training the final dataset on different Convolutional Neural Network (CNN) architectures to classify into Normal case and three disease conditions. A best case accuracy of 92.06% has been obtained by using a modified EfficientNet architecture.