Cortex

The business

Cortex is a clinical AI product that predicts patient deterioration from vitals and history, built under the Google Student Ambassador Program.

Its site had to do something unusual for a product page: make a risk-prediction model legible to the clinicians who would have to trust it, before making any claim about how good it is.

The challenge

Predictive clinical software has a credibility problem before it has a marketing problem. Clinicians will not act on a risk score they cannot interrogate, so a page that leads with accuracy claims argues the wrong point. The site had to show what the model weighs and how a prediction turns into a next step.

The approach

The interface leads and the copy follows. A dashboard preview sits beside the headline showing feature importance, a live risk score, and suggested next steps, so the explainability argument is visible before it is stated. The page is then structured as an argument — the failure of reactive care first, the product second.

What we worked on

• Dashboard demo
  • Feature-importance chart
  • AI analytics insights
  • Risk-score gauge
  • Heart rate and SpO2 readings
• Clinical workflow
  • Suggested next steps panel
  • Four-step onboarding walkthrough
• Product site
  • Four-card problem framing
  • Feature set section
  • In-action section
  • FAQ on accuracy, inputs, speed and safety
• Visual system
  • Dark clinical interface in deep green
  • Glass dashboard panel
• Build
  • React and Vite front end
  • React Router
  • Deployed on Vercel

The solution

A dark clinical interface in deep green, anchored by an interactive dashboard preview. Below it, a four-part framing of the problem, a four-step onboarding walkthrough, the feature set, an in-action section, and an FAQ that takes the hard questions directly.

Business value delivered

A product site that makes an AI prediction interpretable instead of merely impressive - credibility built by showing the working, not by asserting accuracy.

• Predictions that are easier to interpret at a glance
• Stronger product credibility with a sceptical clinical audience
• A clear picture of the workflow from data to decision
• The hard questions - accuracy, inputs, safety - addressed on the page