SalajPro/PneumoX ? reverse-engineered prompt

Build me a small Python project for chest X ray pneumonia classification based on the RSNA Pneumonia Detection Challenge.

I want a simple end to end workflow that can download and prepare the dataset from Kaggle, train an EfficientNetB0 model with basic data augmentation and checkpoint saving, then evaluate it with useful plots like a confusion matrix, ROC curve, precision recall curve, and a classification report. Please also make a small desktop app where I can load a single X ray image, get a prediction, and see a Grad CAM overlay so I can understand what the model is looking at.

Keep it practical and easy to run from the command line with separate scripts for data prep, training, testing, and the app. Use Python and TensorFlow, and make sure the code is organized so it can run on a normal laptop. If anything about the dataset or Kaggle setup needs to be checked online, look up the current docs and follow that.