Project Name: MEDLI Project Title: "Know which doctor to see before you go" MEDLI-makes your hospital visit easier Project Description:
Inspiration Every time someone in India feels unwell, the first question is "which doctor do I go to?" People guess, go to the wrong specialist, waste money and time. We wanted to fix that with AI.
What it does MEDLI is an AI-powered hospital visit assistant. It lets patients describe their symptoms in English or Hindi, then uses Google Gemini to triage and recommend the right specialist with urgency level. It also lets you book doctors, book lab tests with AI explanations, and read prescriptions using Gemini Vision.
How we built it Built with React, Vite, and Tailwind CSS on the frontend. All AI features use Google Gemini 2.5 Flash API — multi-turn chat for symptom triage, structured output parsing for triage summaries, and Gemini Vision for prescription reading. Deployed on Vercel.
Challenges we ran into API quota limits on free tier were exhausted during testing. Venue WiFi was blocking Google API domains, so we had to switch to mobile hotspot. Setting up the correct Gemini model version took multiple attempts.
Accomplishments that we're proud of Built a fully working AI triage system that supports Hindi and English. The pre-visit summary feature — where Gemini generates a clinical note patients can hand their doctor — is something no existing app does.
What we learned How to use Gemini's multi-turn chat API, structured prompt engineering to extract triage data, and Gemini Vision for document reading.
What's next for MEDLI Firebase authentication for patient records, real doctor database integration, appointment reminders, and expanding language support to Tamil, Bengali, and Marathi.
Accomplishments that we're proud of Built a fully working AI triage system that supports Hindi and English. The pre-visit summary feature — where Gemini generates a clinical note patients can hand their doctor — is something no existing app does.
What we learned How to use Gemini's multi-turn chat API, structured prompt engineering to extract triage data, and Gemini Vision for document reading
What's next for MEDLI Firebase authentication for patient records, real doctor database integration, appointment reminders, and expanding language support to Tamil, Bengali, and Marathi.
Tech Stack: css gemini html5 react vercel vite
Log in or sign up for Devpost to join the conversation.