https://www.loom.com/share/49cba8d0ff524b8bab0c62324a7b1d6a

HealthAI Assistant is a full-stack, AI-powered healthcare ecosystem designed to streamline early-stage patient care by combining multimodal AI, predictive diagnostics, geospatial intelligence, and automated teleconsultation workflows into a single production-grade platform.

This project demonstrates advanced AI system design, full-stack orchestration, and real-world healthcare problem solving.

• Multimodal AI core (text + vision)
• Large Language Models for clinical reasoning
• Automated telemedicine & payment workflow
• Real-time hospital discovery using geospatial data
• End-to-end AI + backend + frontend integration

• Implemented Llama 4 Scout (17B Vision) to process unstructured medical PDFs, lab reports, and scanned documents.
• The model reads and understands:
  • Lab values
  • Medical terminology
  • Diagnostic indicators
• Automatically extracts critical patient information from visual and textual medical data.

This enables the system to “read reports” and “see medical documents” like a clinician.

• Integrated Llama 3.3 (70B) as the conversational and reasoning engine.
• Provides:
  • Symptom-based disease prediction
  • Personalized recovery and precaution plans
  • Context-aware medical guidance
• High-parameter architecture allows nuanced medical reasoning, outperforming smaller models in complex health scenarios.

• Built using Python + FastAPI
• Handles:
  • Model inference
  • Workflow automation
  • API orchestration
  • Secure request handling

• Integrated Leaflet.js with Overpass API
• Uses live geolocation to:
  • Identify nearby hospitals
  • Display facilities on an interactive map
• Bridges the gap between diagnosis and treatment access.

Designed a complete business logic loop:

1. User receives AI-based diagnosis
2. System generates UPI QR code for payment simulation
3. After confirmation:
   • Google Meet link is dynamically created
   • Appointment details are auto-generated
4. Confirmation is sent via SMTP email automation

This workflow demonstrates understanding of monetizable, real-world healthcare systems.

• Designed structured, safety-aware prompts for:
  • Llama 3.3 (70B)
  • Llama 4 Scout (17B Vision)
• Focused on:
  • Clinical accuracy
  • Reduced hallucinations
  • High-stakes healthcare reliability

• Backend: Python, FastAPI
• Frontend: JavaScript, Leaflet.js
• AI Models: Llama 3.3 70B, Llama 4 Scout 17B Vision
• APIs:
  • OpenAI / LLM APIs
  • Overpass API
  • Google Meet API
  • Google Translate
  • SMTP (Email Automation)

Backend

• Python
• FastAPI

AI / ML

• Llama 3.3 (70B)
• Llama 4 Scout (17B Vision)
• Scikit-learn
• Pandas
• NumPy

OCR & Document Processing

• pdfplumber
• pdf2image
• pytesseract

Frontend & Mapping

• JavaScript
• Leaflet.js
• Overpass API

Utilities & Services

• OpenAI API
• Google Meet API
• SMTP
• qrcode
• uuid
• python-dotenv

Role: AI Engineer & Full Stack Developer

• Multimodal AI and API Integration
• Agentic Workflow Design
• Chatbot Integration
• Medical Report Analyzer Design
• Backend & full-stack orchestration

📧 Email: [email protected] 🌐 GitHub: https://github.com/OS264

Role: AI Engineer & System Architect

• Multimodal AI integration
• LLM-based clinical reasoning
• System architecture design
• Telemedicine workflow automation
• Backend & full-stack orchestration

📧 Email: [email protected]
🌐 GitHub: https://github.com/shohebattar428

• ✅ Full Source Code
• ✅ PPT Presentation
• ✅ Project Report
• ✅ Technical Documentation

📩 For access:
Email: [email protected]

⭐ If you find this project valuable, consider starring the repository!