Audhd.io

A resource-matching platform for the neurodivergent community: four screeners that score entirely on-device, a state-specific provider directory, and a content engine that cut manual overhead by 90%.

Role

Architect and engineer

Period

Aug 2024 — Mar 2026

Stack

Next.js App Router · TypeScript · Supabase · Tailwind · n8n

Outcome

Zero answers leave the device

The Audhd.io homepage, opening on the first screener question so visitors see what they are walking into.

A resource-matching platform for the neurodivergent community, built around one constraint that shaped every other decision: screener answers never leave the device.

The privacy constraint

Self-screening for ADHD or autism means answering questions people have not said out loud to anyone. Sending that to a server — even a server I control, even encrypted — changes what the tool is.

So scoring and persistence both run client-side. Questions, scoring logic, and saved progress live in the browser; results are written to localStorage and never transmitted. There is no submission endpoint to breach, no answer table to leak, and no subpoena-able record. The privacy property is structural rather than promised — it holds because the data has nowhere to go, not because a policy says it will be handled carefully.

Four screeners run on that model: ADHD, autism, AuDHD, and burnout.

The rest of the platform

  • A state-specific provider directory, because “find a provider” has fifty different answers depending on where you live — licensure, coverage, and waitlists are all jurisdictional.
  • An affiliate storefront with a tracked redirect layer, so the site can fund itself without putting a paywall in front of people looking for help.
  • A content engine on n8n that cut manual content work by roughly 90%.

Accessibility as the product

For this audience, accessibility is not a compliance checkbox — the people most likely to use the site are the people most likely to bounce off a badly built one. WCAG conformance is covered by a dedicated test suite rather than a one-time audit: the screener, blog, and landing flows each have accessibility tests that run in CI, so a regression fails the build instead of quietly shipping.

← All work