medledger-ai

• 

  this is the main landing page

• 

  patient login page and registration page

• 

  doctor login page and registration page

• 

  doctor dashboard

• 

  create a medical record and parse the EHR data file as per existing or new user and save save it in record

• 

  get a quick history of patient with ai summary from records

Inspiration

Healthcare data is often fragmented across PDFs, EHR systems, and insurance portals. Fraud, duplicate claims, and inconsistencies go unnoticed until financial or clinical damage occurs. We were inspired to build a proactive system that doesn’t just store medical records — but intelligently analyzes and secures them in real time.

What it does

MedLedger AI is a secure, AI-powered healthcare intelligence platform. It allows users to upload medical records (PDFs), extracts structured data, summarizes patient history, and detects anomalies or suspicious patterns in real time.

Instead of being a passive EHR system, it acts as an intelligent monitoring layer over healthcare data.

How I built it

Frontend: Next.js 14 (App Router), TypeScript, Tailwind CSS

Backend: Next.js Server Actions with Supabase (PostgreSQL)

Authentication: JWT session cookies using jose + bcrypt

Email: Resend for OTP verification and password resets

AI: Vercel AI SDK with OpenAI (gpt-4o-mini) or Gemini (gemini-2.0-flash)

PDF Parsing: unpdf for extracting medical document text

The system processes uploaded PDFs, sends extracted content to the AI layer for summarization and anomaly detection, and stores structured results securely in the database.

Challenges I ran into

parsing inconsistent and unstructured medical PDFs

Managing AI API rate limits and token costs

Designing anomaly detection logic without labeled fraud datasets

Securing session-based authentication properly

Maintaining fast performance while processing large files

Accomplishments that I'm proud of

Built a full-stack AI healthcare platform from scratch

Implemented secure JWT-based authentication

Integrated real-time AI summarization and chat over medical records

Designed a scalable PostgreSQL schema for structured healthcare data

Successfully combined AI + security into a working prototype

What I learned

How to architect AI-driven applications securely

Efficient API usage and cost optimization strategies

Handling structured + unstructured data pipelines

Implementing secure authentication in server-side Next.js

Building scalable, production-style project architecture

What's next for Untitled

Add real-time fraud risk scoring models

Implement role-based dashboards (provider, auditor, insurer)

Add blockchain-based audit logging for record immutability

Improve anomaly detection using Isolation Forest or ML models

Prepare for HIPAA-compliant deployment and real-world pilot testing

Built With

• and
• bcrypt-for-password-hashing-email-services:-resend-(otp-verification-&-password-reset)-ai-layer:-vercel-ai-sdk-with-openai-(gpt-4o-mini)-or-google-gemini-(gemini-2.0-flash)-?-used-for-pdf-extraction
• chat
• document
• for
• frontend:-next.js-14-(app-router)
• medical
• medical-history-summarization
• parsing:
• pdf
• supabase-(postgresql)-authentication:-session-based-jwt-(via-jose)
• tailwind-css-backend:-next.js-server-actions
• text
• typescript
• unpdf