Skip to content

Co-founder · Forward-Deployed Engineer

MedaLink

An AI front desk for clinics - replies, scheduling and recalls in one system.

Messages from LINE, Facebook, Instagram and the website land in one inbox. The AI answers by each clinic's own rules, books people into slots where the doctor, room and equipment are all free, then takes over reminders, rescheduling, aftercare and six-month recalls.

Live2026.07 – now

MedaLink
2 people
Requirements to production
4 channels
Messages in one inbox

What it does

  1. Only bookable slots are offered

    Only bookable slots are offered

    A treatment is split into stages, and a slot only appears when every doctor, assistant, room and machine each stage needs is free. The front desk stops phoning people to fix bad bookings.

  2. Every patient's stage at a glance

    Every patient's stage at a glance

    Bookings create a record automatically and move through a seven-stage pipeline; reminders, rescheduling, waitlist fills, aftercare and six-month recalls can each be switched on or off.

  3. What must not be said, is not said

    What must not be said, is not said

    Promises of results, comparative claims and post-treatment symptom advice are on a do-not-answer list - the AI stops and alerts the front desk. Wound and medical photos always go to a human.

Try it: a time that actually works

A treatment is a chain of steps, and each step needs its own people, room and machine. Pick one: only the times when all of them are free come back.

Treatment

2 times fit today

A made-up day at a made-up clinic. The rule is the real one.

More than auto-replies

A front desk can answer messages for about 176 hours a month; for the other 544, messages keep arriving. But auto-replies are not enough - once the question is answered, the person has to be booked into a slot that really works, followed by reminders, changes and recalls. MedaLink connects that whole chain.

What I did

Co-founder · Forward-Deployed Engineer

  • The AI half of a two-person team, from the conversation engine to cloud deployment
  • Made the AI answer by each clinic's own playbook, knowledge base and do-not-answer list, in the clinic's own voice
  • Routed messages to different models by difficulty, so simple greetings use a cheaper model and save the clinic money

Built with

React 18 · Vite 6 · GSAP · Recharts · Cloudflare Pages Functions · jose · LLM Multi-Agent · Cloudflare D1 · Cloudflare R2 · Cloudflare Pages · Cloudflare Workers Cron · Browser Rendering · Wrangler