Skip to content

Medical AI · Healthcare Technology

Chest Doctor: AI-Powered Medical Diagnosis Platform

Full-stack medical AI application combining DenseNet-201 chest X-ray classification (95% accuracy) with multi-agent RAG chatbot for healthcare professionals.

Role
Full Stack AI Developer
Built end-to-end medical AI platform from deep learning models to production deployment.
Industry
Healthcare Technology
Timeline
Completed

Business impact

accuracy
95%
inference speed
60 images/sec
medical documents
60+
deployment
Azure GPU

Overview

Built a comprehensive medical AI platform for chest disease diagnosis, integrating deep learning image analysis with conversational AI to assist healthcare professionals in diagnostic workflows.

The platform combines computer vision for X-ray analysis with a knowledge-augmented chatbot trained on 60+ medical documents, enabling doctors to get AI-assisted diagnoses with explainable results through Grad-CAM visualizations.

Technical architecture

ML Pipeline
DenseNet-201, ONNX INT8 quantization, PyTorch, Grad-CAM
Knowledge Pipeline
PyMuPDF extraction, Azure OpenAI embeddings, HNSW indexing
AI Layer
GPT-4o-mini, LangChain, LangGraph, Azure AI Search
Backend
FastAPI, Azure Cosmos DB (MongoDB), JWT auth
Frontend
React, TypeScript, Vite
DevOps
GitHub Actions, Azure Container Apps (GPU), Docker

Key features

  • Chest X-ray classification using DenseNet-201 with 95% accuracy
  • Grad-CAM explainability for transparent AI decisions
  • GPU-accelerated inference processing ~60 images/sec
  • Multi-agent RAG chatbot with intent routing
  • Knowledge base from 60+ medical documents
  • Hybrid vector + keyword search for accurate retrieval
  • Doctor-patient appointment booking system
  • JWT authentication for secure access

Challenges solved

  • Achieving 95% accuracy on medical image classification
  • Implementing Grad-CAM for explainable AI in healthcare context
  • Building multi-agent RAG with intent routing and query rewriting
  • Optimizing GPU inference for real-time processing

Tech stack

  • Python
  • PyTorch
  • DenseNet-201
  • ONNX
  • FastAPI
  • LangChain
  • LangGraph
  • RAG
  • Azure AI Search
  • Azure OpenAI
  • React
  • TypeScript
  • Docker
  • Azure Container Apps