Peer-reviewed methodology. Calibrated by consultant plastic surgeons.

Ruxi Vision
Skin analysis built on clinical cases

A skin-analysis vision model calibrated by consultant plastic surgeons on real clinical cases. It scores visible skin concerns on 1–5 clinical scales, and is offered as an API and as a white-label scanner that clinics and skincare brands embed on their own websites.

Training data
6,500
clinician-annotated facial photographs
Coverage
30 cells
6 Fitzpatrick phototypes × 5 age bands
Model accuracy
MAE 0.6
mean absolute error on the 1–5 clinical scale

What Ruxi Vision does

One photograph in, a set of clinical scores and a readable explanation out. The scores come from the vision model; the words are written around them, never the other way round.

Scores on clinical scales

Fourteen facial metrics, including regional wrinkles, laxity, pigmentation, redness, blemishes and pores, each scored on the same 1–5 clinical scale the annotating surgeons used.

Narrative around the scores

Once the vision scores are fixed, a language model writes the explanation, an estimated skin age and a skincare routine. It is told the scores and not allowed to change them.

API or embedded scanner

Call it from your own software, or embed a branded scanner on your website with one iframe tag. The scanner shows your products and needs no patient accounts.

How the model was built

Ruxi Vision is a hybrid pipeline. The part that scores skin is our own vision model, trained on a clinician-annotated dataset. The part that writes is a language model that works from those scores.

  1. 1

    A clinician-annotated dataset

    6,500 facial photographs, collected with UK GDPR-compliant consent and annotated by clinicians. The dataset is structured into 30 demographic cells: 6 Fitzpatrick phototypes across 5 age bands.

  2. 2

    Surgeon calibration

    Consultant plastic surgeons graded each concern on 1–5 clinical scales. Agreement between graders was measured before training: Krippendorff's alpha 0.75–0.83 and ICC 0.78–0.86 across the scales.

  3. 3

    The vision model

    A vision model trained to reproduce those grades from the photograph alone. On the 1–5 clinical scale its mean absolute error is 0.6. The scores it returns are deterministic for a given image.

  4. 4

    Narrative generated around the scores

    The fixed scores are handed to a language model with our clinical prompts. It produces the explanation, estimated skin age, recommendations and routine. We re-apply the vision scores over its output before anything is returned.

What we claim, and what we do not

Peer-reviewed methodology. The dataset construction and the inter-rater reliability of the grading are published (see Research below).

Calibrated by consultant plastic surgeons. The grades the model learns from were produced by consultant plastic surgeons, not crowd-sourced labels.

Clinically validated scoring. Model accuracy is reported as mean absolute error on the 1–5 clinical scale. Papers on the training process and on dermatologist validation of the scoring are in preparation.

Not a medical device. Ruxi Vision assesses visible skin quality for cosmetic purposes. It does not diagnose, and every output carries that statement.

Clinical validation and research

The dataset and grading method behind Ruxi Vision are published in a peer-reviewed journal.

Paper 1 · published

Development and reliability assessment of a clinician-annotated facial skin dataset for AI-driven aesthetic analysis: methodology and inter-rater reliability

Gilleard O, Simsek A, Alamouti R, Sadri A.

European Journal of Plastic Surgery. 2026;49:179. DOI 10.1007/s00238-026-02561-0

  • 6,500 clinician-annotated facial photographs
  • 30 demographic cells: 6 Fitzpatrick phototypes × 5 age bands
  • Inter-rater reliability: Krippendorff's alpha 0.75–0.83, ICC 0.78–0.86
  • UK GDPR-compliant consent
Read the paper at Springer

Paper 2 (training process): in preparation.

Paper 3 (dermatologist validation of scoring): in preparation.

Embedded scanner

Your branding, your products, one iframe tag

A skin scanner that lives on your own website. Visitors take a photo, get their scores and a routine built from the products you sell, and can buy or book without leaving your page.

  • Your logo, colours and font, edited in your Ruxi.AI dashboard
  • Your product catalogue: routines are built only from what you stock
  • No patient accounts. Visitors scan as guests; nothing to sign up for
  • Optional treatments section and WhatsApp booking for clinics
  • In-app checkout or hand-off to your existing cart
  • Your API key stays on our servers; the browser never sees it

The whole integration

<iframe
  src="https://embed.skincognition.com/e/your-clinic/"
  allow="camera"
  style="width:100%;max-width:480px;height:85vh;min-height:700px;border:0"
  title="Your Clinic skin analysis">
</iframe>

We connect your shop (WooCommerce or Shopify) and payment during set-up. After that, branding and wording changes you make in the dashboard appear in the scanner within a minute.

How a scan works

The same four steps whether you call the API or use the embedded scanner.

01

Capture

One front-facing photograph in even light, taken in the browser or uploaded.

02

Score

Ruxi Vision returns fourteen clinical scores on the 1–5 scale, plus the model version that produced them.

03

Explain

A language model writes the explanation and estimated skin age around the fixed scores.

04

Recommend

A skincare routine built from the products you supply, with optional treatment options for clinics.

Pricing

Ruxi Vision is priced per project rather than from a public rate card. Three parts make up a quote:

One-time set-up fee

For a branded scanner: branding, shop connection, payment set-up and QA on your site.

Annual platform fee

Covers the dashboard, the embed hosting, the model updates and support.

Scan-volume package

Choose a yearly scan allowance that fits your traffic. Packages scale from small clinics to large brands.

Quotes on request. A one-month API trial is available after sign-up.

Also built on Ruxi Vision

SkinCognition is our own consumer skin-scanner app for iOS, Android and the web. It is the same scanner that powers the embedded version.

Important: Ruxi Vision is not a medical device and does not diagnose any condition. It assesses visible skin quality from a photograph for cosmetic and educational purposes. Scores are produced by our own vision model, trained on a clinician-annotated dataset and calibrated by consultant plastic surgeons; the accompanying text is generated by a language model from those scores. Anyone with a skin concern should consult a qualified clinician.

Talk to us

Clinics and skincare brands: the founder handles access requests and quotes directly. Developers with integration questions can reach the engineering side.