Skin diseases are common, and identifying them manually is slow and depends on scarce dermatological expertise. Existing automated systems also tend to be sensitive to image quality, act as black boxes that give no reason for their predictions, and stop at diagnosis without helping users act on the result. This paper presents an AI-powered dermatology system that detects skin conditions from images using a convolutional neural network (CNN) and explains each prediction using Gradient-weighted Class Activation Mapping (Grad-CAM). Every prediction is accompanied by a confidence score, and the Grad-CAM heatmap highlights the image regions that influenced the decision, which improves transparency and trust. The detected condition is then used to recommend suitable skincare products, and the system compares the same or similar products across brands and shopping websites by price and availability. The pipeline combines image preprocessing and capture guidance, a TensorFlow/Keras CNN, and a FastAPI backend, giving the user a prediction, an explanation and an affordable product choice in a single flow. The system is intended as an AI-assisted screening and support tool and not as a replacement for professional medical diagnosis. Future work includes multilingual and audio-based support.

CNN detection, Grad-CAM, recommendations