MediScope AI

• 

• 

• 

• 

• 

• 

• 

• 

• 

Inspiration

Modern healthcare demands faster, smarter decisions—especially in critical scenarios. We envisioned a unified AI system that helps clinicians triage patients, extract insights from handwritten notes, and centralize case information in real time.

What it does

MediScope AI is a full-stack Clinical Decision Support System (CDSS) that enables:

AI-based patient triage with urgency detection

OCR + NLP-driven medical report scanning

Structured dashboards for clinicians to monitor patient cases

How we built it

Frontend: React + Vite for fast, responsive UI

Backend: Node.js + Express API for routing and MongoDB for storage

ML Models: Python FastAPI microservices for LLM integration using Ollama (Phi-3), OCR with EasyOCR

LLM: Phi-3 via llama-cpp-python for both triage and report parsing

Docker: Containerized services for scalable deployment

Challenges we ran into

Integrating local LLMs with FastAPI in a performant way

Parsing clean JSON from raw LLM output reliably

Ensuring frontend and backend communication across services

OCR quality on handwritten reports

Accomplishments that we're proud of

A working end-to-end triage + report scanner pipeline

Seamless integration of LLMs with custom prompts

Dynamic clinician dashboard with real-time updates

No dependency on cloud LLM APIs—fully local AI stack

What we learned

Fine-tuning prompt engineering for clinical use cases

Bridging frontend-backend-LLM workflows smoothly

Optimizing performance and error handling across async microservices

What's next for MediScope AI

Add EHR integration for hospital systems

Support multilingual medical documents

Enable case timeline view and real-time collaboration

Add audio transcription for doctor dictation

Deploy on a secure cloud VM for real clinical trials

Built With

• css
• docker
• easyocr
• express.js
• fastapi
• javascript
• llama-cpp-python
• mongodb
• node.js
• ollama
• python
• react
• tailwind
• vite