Glowniq: An AI Skin-Analysis App Built Around a Selfie
Glowniq turns a selfie into personalized skin scores and a simple AM/PM routine. Here is the multi-provider AI fallback architecture, the credit-based billing model, and the GDPR groundwork behind a consumer skincare app built by a solo founder.
A selfie is the input. Skin scores and a two-step routine are the output. Everything Glowniq does sits between those two points — and most of what's hard about the product lives in the part users never see: which AI model actually answered the request, and whether the answer can be handled safely under GDPR when the input is a photo of someone's face.
The product surface
One flow, kept deliberately narrow:
- Capture. The user takes or uploads a selfie in the mobile or web app — no salon, no appointment, no human in the loop.
- Analysis. The photo is scored across oiliness/shine, redness, texture, pores, and a handful of other signals, then mapped to a skin type.
- Routine. The output is a simple AM/PM skincare routine, not a diagnosis — Glowniq is explicit that it doesn't replace a dermatologist.
Why the AI layer has three fallbacks, not one
A single-provider AI product has a single point of failure: a rate limit, an outage, or a model deprecation, and the core feature is down. Glowniq routes skin-analysis requests through OpenRouter first, with direct OpenAI, Anthropic, and Groq integrations as fallbacks if the primary path fails or times out. The user never sees which provider answered — they see a score. That redundancy is boring infrastructure, but it's the difference between "the app is down" and "the app is a little slower today."
Billing without a sales team
There's no salon, no seat count, no invoice to negotiate — it's a solo consumer app, so the billing model is self-serve by design:
- Free tier. Three credits to try the analysis before paying anything.
- Starter / Pro / Premium. Credit-based paid tiers, billed through Stripe, with a discount on the yearly plan over monthly.
- No B2B pricing. Every user is a consumer, and every credit is spent the same way regardless of who's asking.
The part that doesn't show up in a screenshot: GDPR
A face photo sits close enough to biometric data that the compliance work isn't optional polish — it ran alongside the product work, not after it: a subprocessor register, a breach-response workflow, and a DPIA (data protection impact assessment), specifically because the core feature is "upload a photo of your face."
Where it stands
The web app is live in production. The mobile apps are earlier in their rollout — Android is in Play Console internal testing, and iOS is working through TestFlight — so today's mobile audience is testers, not the public app stores yet. The next milestone is exactly that: getting both builds past internal testing and into general release.