I need a web application that lets visitors upload or snap a photo of their skin and receive an instant, AI-generated diagnosis of visible conditions. The scope is focused and image-only: no forms, no personal health questionnaires, just computer-vision analysis that can reliably flag issues such as acne, eczema, rosacea, suspicious moles, or other dermatological concerns. The flow I have in mind is straightforward: a clean landing page, an image-capture or upload component, a secure backend that feeds the picture through a convolutional-neural-network model, and a results page that names the detected skin condition, shows confidence scores, and offers a short, plain-language description of what that condition means. An admin area where I can review anonymised images, correct labels, and trigger model retraining would be a bonus. You may build the model from scratch in Python using TensorFlow or PyTorch, or integrate an existing dermatology API if accuracy is comparable. The frontend can be React, Vue, or another modern framework so long as it’s mobile-friendly; users are likely to be on their phones. Minimum acceptance criteria • Users can upload a JPG/PNG or capture a live photo. • The system returns a labelled diagnosis for at least the five most common skin conditions with >85 % top-1 accuracy on my validation set. • All processing happens server-side; no images are stored once the result is delivered unless I opt in via the admin panel. • Basic rate-limiting and HTTPS enabled for security. • Code, trained weights, and deployment instructions delivered in a private repo. Future phases may expand to skin-type classification or product recommendations, so structure the code to allow additional endpoints later.