
DOCUMENT INTELLIGENCE / MEDICAL EDUCATIONLIVE ON AZURE
Medical RAG Chatbotby Syed Tabish Shah.
A source-grounded question-answering application that retrieves relevant passages from trusted WHO medical documents before generating an educational response with source pages.
Medical RAG interface deployed through Docker and Azure Container Apps.
01 / PROJECT OVERVIEW
Grounding answers in medical sources.
The project explores retrieval-augmented generation for medical education. Instead of asking a language model to answer from general memory alone, the application searches an indexed collection of WHO reference documents and supplies the most relevant passages as context.
The original internship implementation used Oracle AI Database 26ai for dense vector storage and Ollama embeddings. The live cloud version uses Gemini embeddings, Azure AI Search, and Gemini answer generation while preserving source-page references.
KNOWLEDGE BASETrusted WHO reference PDFs
TOPICSAnemia, diabetes, and hypertension
DEPLOYMENTDocker on Azure Container Apps
02 / RETRIEVAL FLOW
From question to grounded response.
User questionThe application receives a medical education query.
Vector searchAzure AI Search finds relevant document passages.
Context assemblyRetrieved passages and source pages become model context.
Answer generationGemini produces a response grounded in retrieved material.
03 / TECHNOLOGY
Two implementations, one retrieval idea.
PythonRAGOracle 26aiOllamaAzure AI SearchGeminiFastAPIStreamlitDockerAzure Container Apps
This project connected document processing, chunking, embeddings, vector similarity search, prompt construction, API integration, containerization, and cloud deployment in one end-to-end application.
Medical-use notice
This application provides general educational information. It does not replace professional medical advice, diagnosis, or treatment.