← All work

Senzone

Film your walk. Get the right insoles.

Senzone helps people with foot pain. I built the platform that turns a phone video of someone walking into an expert-grade foot analysis and an insole recommendation — with Senzone’s own specialists teaching the AI along the way.

My role
Sole developer
When
2024 — 2026
See it live
senzone.dk

Foot pain is common, and the right insole can make a real difference. The hard part is getting good advice. Senzone wanted that advice to be one short video away — and just as good as asking one of their experts.

How it works

  1. 01

    A foot analysis, from your sofa

    Getting the right insoles usually means a trip to a specialist. Senzone wanted anyone with a phone to get the same advice — without losing the expertise that makes it good.

  2. 02

    Film ten seconds of walking

    The customer films themselves walking and uploads the video in the app. That’s all they have to do.

  3. 03

    The AI watches — and looks things up

    The AI studies the video frame by frame. Before it answers, it looks up the five most similar cases Senzone’s experts have already approved, so it answers the way they would.

  4. 04

    Experts keep it honest

    Every answer can be rated and corrected by Senzone’s foot specialists. The ones they approve go into the library — so the AI gets better every day they work.

  5. 05

    From advice to doorstep

    The recommendation is always a real insole from Senzone’s range. One click, and it’s paid for and on its way.

videoanalysesimilar casesorderapproveThe customerfilms their walkThe appupload & orderThe engineruns the analysisFoot expertscheck the AIExpert workbenchrate & correctThe AIwatches the videoCase libraryapproved answersPayment & shippingQuickPay · Shipmondo
  • Built by me
  • Outside service
  • People

Choices I made

  1. 01

    The AI never works alone

    A wrong recommendation costs the customer money and comfort — and Senzone its reputation.

    Experts rate and correct the AI’s answers, and only their approved cases are used as examples.

    The catch It takes expert time — which is exactly what makes the answers trustworthy.

  2. 02

    Learn from the experts, every day

    An AI usually only improves when someone rebuilds it, which is slow and expensive.

    Every case the experts approve becomes an example the AI consults next time.

    The catch The answers are only as good as the library — so the workbench is built to make growing it quick.

  3. 03

    Only recommend what’s on the shelf

    A general AI will happily invent a product that doesn’t exist.

    Recommendations are limited to Senzone’s own 70+ insoles, so every answer can be ordered.

    The catch A new product has to be added to the range before the AI can suggest it.

The result

Under the hood — for the technical reader

Stack Laravel · Filament · Next.js · Azure OpenAI · Azure AI Search

  • AI foot analysis 2024

    Video frames analysed by a vision model, grounded in similar approved cases and matched to the insole range, with checkout built in.

    Laravel · Next.js · Azure OpenAI · QuickPay · Shipmondo

  • Expert workbench 2025

    Experts rate and correct AI answers in real time; approved cases are embedded and used to ground the next analysis.

    Next.js · Laravel Reverb · Azure AI Search