destryet/datathon_ocr_medical_bills

AI Medical Bill Extractor (HackRX Datathon)

This is a High-Performance API for extracting line items from complex medical bills using Deep Learning.

🚀 Architecture

• —Engine: PaddleOCR (Industrial Grade OCR) for text detection.
• —Backend: FastAPI (Asynchronous Python Framework).
• —Optimization: "Lazy Loading" architecture with self-healing dependency management to ensure 99.9% uptime on serverless environments.
• —Logic: Custom parsing algorithms to detect tabular structures and filter out noise.

🛠️ Tech Stack

• —Python 3.10
• —FastAPI / Uvicorn
• —PaddleOCR / OpenCV (Headless)
• —Numpy / Pandas

🔗 API Endpoint

Base URL: https://destryet-datathon-ocr-medical-bills.hf.space

Endpoint: /extract-bill-data

• —Method: POST
• —Body: {"document": "URL_TO_PDF"}