Medical Bridge

Built in 16 hours at the Batam Singapore Hackathon 2026 (Health Tourism track) as team Running on Coffee, where it placed 2nd Runner-Up.

The problem

Batam is under an hour from Singapore by ferry, and treatment there costs less. What stops patients is the logistics: hospital, specialist, ferry, hotel and transfers are five separate bookings, and they depend on each other. The hotel has to be near the hospital, and the ferry home has to fall after the recovery period.

What it does

• Patient web chat: the patient says what they need, taps through a few questions (date, party size, budget, nights) and gets an itemised plan they can edit. Swap the hospital or drop the hotel, and the total updates as they go.
• Operations dashboard: hospital staff see every request, check each line of the quote, then approve or reject it.
• Partner portals: hospitals, hotels, ferry and transport operators each see their own bookings and keep their own rates up to date.
• Patient pass: once approved, the patient gets a private link with the whole trip, step by step, on their phone.

The AI never gets the last word

• The model only fills in the blanks (treatment, date, party size, budget) as structured JSON. Every sentence the patient reads comes from a fixed question bank, so nothing the model writes reaches the screen.
• Prices come from the catalogue, never from the model.
• A quote reaches the patient only after a named staff member approves it. Low confidence, high-risk procedures, unknown treatments and emergency language always go to a person.
• If the model is down, a keyword matcher takes over and the case goes to a person, so no patient gets dropped.

Built with

Laravel 10 API, React 18 with TypeScript, Tailwind, and Gemini through an OpenAI-compatible endpoint. 106 automated tests cover the approval gate, pricing and patient privacy.

The live demo uses real Batam hospitals and hotels (checked against OpenStreetMap) with sample prices, doctors and patients.