Inspiration
Skincare advice is everywhere, but most of it feels generic, confusing, or product-driven. We wanted to build something more personal: a coach that helps people understand their own skin signals and track improvement over time.
What it does
Skinwise lets users upload or capture a selfie, analyzes visible cosmetic skin signals, and turns the results into plain-language insights. It generates a personalized skincare routine, offers an AI chat coach, and stores past results so users can follow their transformation.
How we built it
We built Skinwise as a web app with a camera/upload flow, skin analysis API integration, account creation, saved history, and an AI coach. The LLM layer is provider-agnostic, so the app can switch between OpenAI-compatible providers using environment variables.
Challenges we ran into
The hardest parts were handling real photo quality issues, making the camera work smoothly across desktop and mobile, and getting deployment working cleanly on Vercel. We also had to design the AI layer so no feature depended on one specific model provider.
Accomplishments that we're proud of
We’re proud that Skinwise feels like a real product, not just a demo. It has analysis, routines, chat, accounts, history, and progress tracking. We’re also proud of making the AI architecture flexible enough to support different providers without rewriting app logic.
What we learned
We learned a lot about building AI features that are useful only when they are grounded in real user context. We also learned how important UX is for AI apps: the explanation, trust, and flow matter just as much as the model response.
What's next for skinwise
Next, we want to improve progress charts, add smarter weekly summaries, support product scanning, and make the coach more proactive with reminders and routine adjustments based on each user’s history.
Log in or sign up for Devpost to join the conversation.