The stack behind the scan
Vision, profiling, ingredient reasoning and explanation are separate services, so each can be replaced without rewriting the others.
Signal path
Framing, lighting and stillness checks run before a frame is accepted.
Landmark detection and region segmentation turn a frame into measurable zones.
Measurements are merged with stated answers into a comparable profile.
Ranked concerns are traced through the ingredient graph and filtered by sensitivities.
Every recommendation is paired with a plain-language reason before it is returned.
Models are under active development. We make no claim of clinical validation for any component of this pipeline.
Four rules the stack is built around
If the system cannot state why it recommended something, it should not recommend it. The explanation layer is a requirement, not a feature.
Scores are normalized for lighting and framing so two scans taken weeks apart describe the same skin, not the same room.
Vision, profiling, ingredient reasoning and explanation are separate services, so each can be replaced without rewriting the others.
Anything that edges toward diagnosis is routed to a referral message rather than an answer.
Integration surfaces
Partners can embed the full experience or call the services directly. The endpoints below are illustrative of the shape of the API, not a published contract.
Request early API accessSubmit a capture and receive structured regional measurements.
Retrieve a stored skin profile and its scan history.
Return a ranked routine with a reason attached to each step.
Delete a profile and all data derived from it.