Inspiration Skincare shopping can feel like guesswork. A skin scan may return scores, but those numbers alone do not tell someone what they are looking at or what to do next. I built DermaPilot AI to connect a scan to understandable visual results, practical guidance, and product choices that fit a person’s goals and budget. What it does Users can upload a photo or take one with a live camera guide that helps frame and crop their face. DermaPilot sends the image to the YouCam Skin Analysis API, then displays the returned skin metrics and available spotted-area overlays alongside the user’s photo. Signed-in users can revisit saved scans and privately stored snapshots. Users can save their skincare preferences, explore matching catalog products, and ask an AI skincare guide questions about their scan and routine. The guidance is cosmetic and clearly distinguishes a YouCam observation from a possible explanation; it does not present scan scores as a medical diagnosis. How I built it DermaPilot is a responsive Next.js and React app. MediaPipe supports live face framing, YouCam provides skin analysis, and Afribase handles authentication, scan records, preferences, and private photo storage. An OpenAI-powered chat uses saved scan scores and relevant preferences to answer users’ questions without sending their selfie to the language model. Challenges and what I learned Image quality was a major challenge: faces that looked clear to a person could still be too small for analysis. I added framing guidance, cropping, and clearer image-quality feedback. I also worked through private storage permissions and partial failures so a completed scan can retain its scores even if a snapshot cannot be saved. The biggest lesson was that a useful Skin AI product needs more than an API result. People need to see what the result refers to, understand its limits, and have a sensible next step.