Intiaj Islam
←All workCase study 05 / 10

Carnet

AI-Powered Health Companion & Medical Record Ingestion

Intelligent personal health vault that scans physical medical prescriptions, lab reports, and doctor logs using AI OCR, organizing them into a searchable medical timeline.

Client
Medical AI Platform
Field
AI & HealthTech
Stack
Next.js, OpenAI, Firebase, Node.js, Tailwind
Carnet

Overview

Turning paper prescriptions into a health timeline you can search

The problem

Patients struggle with disorganised binders of lab reports, crumpled paper prescriptions, and confusing medical jargon. When seeing a new specialist, crucial historical context is often lost.

What I built

Carnet combines OCR and GPT-4 Vision with encrypted cloud storage. Patients snap a photo of their lab results, and Carnet extracts the key biomarkers, places them on a chronological timeline, and explains what each one means in plain English.

Features

  1. OCR Document Scanning
    01

    OCR Document Scanning

    Instantly extract text from prescriptions, lab results, and paperwork using our vision OCR engine.

  2. Secure File Upload & Vault
    02

    Secure File Upload & Vault

    Upload medical PDFs, images, and health records with client-side encryption and Firebase sync.

  3. Smart Biomarker Tagging
    03

    Smart Biomarker Tagging

    Automatically categorize lab files by detecting cholesterol, glucose, blood pressure, and dates.

  4. Doctor Sharing & Collaboration
    04

    Doctor Sharing & Collaboration

    Share specific records or structured timelines with your healthcare specialist in one click.

  5. Symptom Logging & Timeline
    05

    Symptom Logging & Timeline

    Log daily symptoms with notes and photos, visualizing biomarker progression over time.

  6. Conversational Health AI
    06

    Conversational Health AI

    Get clear, accessible breakdowns of complex medical terminology with our grounded AI companion.

Results

Joined the waitlist in the first 2 weeks
100+Joined the waitlist in the first 2 weeks
Active clinical beta testers & physicians
20+Active clinical beta testers & physicians
Positive user satisfaction rating
99%Positive user satisfaction rating
Average OCR pipeline extraction latency
800msAverage OCR pipeline extraction latency

How it's built

Client
Next.js App Router with responsive mobile-first UI
Backend
Serverless Node.js edge functions with streaming responses
Data
Encrypted Cloud Firestore with role-based security rules
Services
OpenAI GPT-4 Vision, OCR ingestion pipeline, Firebase Storage

What shipped

  1. 01AI OCR extraction pipeline transforming messy photos into structured biomarkers
  2. 02Conversational health assistant for non-diagnostic medical report summaries
  3. 03Zero-knowledge encrypted cloud storage for sensitive patient PDFs
  4. 04Granular permission system for sharing records with healthcare specialists

Next project

AI Junction→

Unified Multi-Model Discovery & Workflow Automation Platform