Oyster is AI infrastructure for beauty, built in Africa and accurate on every skin tone. Our research asks what beauty and skin health become when skin is finally something you can measure, for everyone on earth.
We work with researchers, clinicians, engineers and partners who want skin to be understood. If any of this is yours, join us in any capacity. Write to [email protected].
A Technical & Clinical Analysis of the Oyster Skin Intelligence Engine.
How Oyster analyses deeper skin tones, checks photo quality and matches products. The report also explains its AI model and data practices.
Turn an ordinary camera into a clinical instrument, and read skin at a depth a selfie never reaches.
Reconstruct hyperspectral skin signatures from a commodity phone camera, so clinical grade sensing no longer needs clinical grade hardware.
Extend Oyster's proprietary photon method to map hydration, pigment and vascularity at increasing depth, from a single capture.
Read skin from ambient light and low cost sensors, for the many settings where a camera is not the right instrument.
A small device that grades skin the same way every day, turning single scans into a trend line a person can actually watch.
One model that reads skin across every tone, knows when it is unsure, and never has to leave the person who owns the face.
A single model that reads structure, tone and concern across every Fitzpatrick type, so accuracy on the deepest skin is the standard, not a patch.
Generate faithful synthetic skin across tones and conditions, so the fairness gaps that real world datasets leave open can be closed on purpose.
Run a full scan on the phone itself, so skin data never has to leave the person who took it, even where connectivity is thin.
Uncertainty aware analysis that says clearly when it is not sure, and hands that case to a human instead of guessing.
Watch skin change over years, model how it responds, and route the rare early signal to a real clinician.
Model how a person's skin changes across years, and what actually moves the curve, from a series of scans rather than a single moment.
A living digital twin of a person's skin that simulates how it will respond, before anything is ever put on the face.
Detect the rare pattern that deserves a professional eye and route it to a qualified clinician, without ever posing as a diagnosis.
Bring the skin microbiome into analysis and recommendation, so care works with the skin's own ecology rather than against it.
A routine that reasons about ingredients, adapts to a life in motion, and shows the result before it happens.
A routine that reorders itself before a product runs out and adjusts as the skin responds, quietly, in the background.
Show a plausible result of a routine or a treatment on a person's own face, months ahead, in augmented reality.
Care that adapts to weather, travel, sleep and hormonal cycles, because skin does not live in a laboratory.
A recommendation engine that reasons about actives, interactions and order of use, not just the label on the box.
Measurement at population scale, intelligence that tells brands what to make next, and a field held to open standards.
Run skincare and cosmetic efficacy studies at population scale, with measurement that stays consistent across every participant.
Turn anonymised skin trends into a signal that tells brands what to formulate next, by region and by tone.
Bring reliable analysis to pharmacies and clinics on slow or intermittent connections, so access never depends on bandwidth.
Publish open benchmarks for skin AI accuracy on the deepest tones, so the whole field can be measured against a standard anyone can see.
Researchers, clinicians, engineers, brands and institutions. Bring a dataset, a question, a method or a doubt. There is room here in any capacity.
[email protected]