SkinCair
Three face photos → a personalized, conflict-checked skincare routine.
Jul 2026 – present
SkinCair analyzes three guided face photos — front, partial-left, and partial-right, with a framing oval and lighting hints — to detect visible skin concerns and generate a personalized AM/PM routine. Instead of replacing your whole shelf, it upgrades or adds one product per detected concern, built around what you already own.
Recommendations come from a deterministic engine seeded with a dermatology-evidence framework: AI analyzes the photos and phrases the explanations, but never decides what to recommend — the same input always produces the same auditable routine. Safety rules are enforced as engine invariants, not conventions: every AM routine ends in SPF 30+, conflicting actives are hard-blocked, and pregnancy-safe filtering is built in.
Key features
Guided 3-photo capture
Front and partial-side shots with a framing oval and lighting hints, so analysis starts from consistent input.
Deterministic recommendation engine
AI perceives, but never decides — the same photos always produce the same routine, auditable by version.
Conflict-checked routines
A hard conflict matrix gates every plan: incompatible actives can't ship together, and at most one new potent active is introduced.
One product per concern
Recommendations build around your existing shelf instead of replacing it wholesale.
Safety as invariants
SPF 30+ ends every AM routine, pregnancy rules out retinoids, and dermatologist-referral triggers are enforced by tests.
Built with
Source code for this project is private.