This repository contains two projects:
A microservice-based chatbot system that answers questions about medical services for Israeli HMOs (Maccabi, Meuhedet, and Clalit) based on user-specific information.
Key Features:
- Bilingual support (Hebrew/English with RTL handling)
- Two-phase interaction: information collection followed by personalized Q&A
- Stateless microservice architecture with FastAPI backend and Streamlit frontend
- Integration with Azure OpenAI GPT-4o
- Custom database construction from provided HTML documents for retrieval-augmented generation (RAG) using keyword-based search
- Exhaustively tested with a comprehensive synthetic dataset to ensure reliability and edge case handling
Technologies:
- Azure OpenAI
- FastAPI
- Streamlit
- RAG (custom DB with keyword search)
A document processing system designed to extract specific fields from National Insurance Institute (ביטוח לאומי) forms in both Hebrew and English.
Key Features:
- OCR processing of PDF and image files
- Intelligent field extraction using Azure OpenAI
- Data validation against predefined schemas
- User-friendly Streamlit interface
- Testing helped identify where targeted use of advanced extraction features would significantly improve accuracy
Technologies:
- Azure Document Intelligence
- Azure OpenAI
- Streamlit
- Pydantic
Each project has its own detailed README with specific setup instructions:
For the complete project requirements, refer to assignment.md.