AI-Driven Medical
Chatbot for
Accurate Health
Assistance
Final Year Project - AI & ML
Lab
Group A
Abstract
• Internet and electronic devices
have made medical resources
more accessible.
• Project builds a medical chatbot
using Retrieval Augmented
Generation (RAG) and the Llama
model.
• Large Language Models (LLMs)
provide better context
understanding and accurate
responses compared to traditional
chatbots.
• A general wiki for frequently
queried diseases and symptoms is
used to improve responsiveness
and reduce computational load.
Introduction
• This project uses Llama with
RAG to develop a medical
chatbot designed to assist
medical professionals, not
replace them.
• Reduce the workload of
healthcare professionals by
addressing general inquiries.
• Combat misinformation on
social media and provide
accurate medical knowledge
to the public.
• Chatbots in healthcare have
been shown to enhance
communication and accessibility,
using advanced technologies like
BERT and deep learning for
1. Data Collection and Preprocessing:
- Sources: 'The Gale Encyclopedia of
Medicine,' Cleveland Clinic, patient.info, drugs.com
- Tools: Python libraries (requests,
BeautifulSoup, PyPDF)
2. Chatbot Development:
Methodology
- Fine-Tuning Model
- Integration: Backend (Django), Frontend
(NextJS), Database (SQLite3)
3. Tools: Jira for project management, GitHub for
version control.
Built on a RAG-based LLM application.
Data integrated and language model is
fine-tuned for handling medical queries.
Chat-Bot
Development The “llama-2-7b-
chat.ggmlv3.q4_0.bin” model is loaded
using the CTransformers library.
Chosen for its versatility and
effectiveness
Scraped data is stored in the
Pinecone vector database.
Retrieves relevant
Chat-Bot information using vector
Development similarity search.
(cont)
Combines retrieved context
with user queries for
informed and contextually
appropriate responses.
Domain Awareness and Query
Handling
• Domain Awareness:
Trained to respond to medical queries,
ensuring relevance and accuracy within its
knowledge domain.
• Out-of-Scope Handling:
The chatbot politely responds that it is
not equipped to handle such queries.
• Importance:
Enhances user trust by avoiding
misinformation and maintains the credibility.
System Design
• 1. System Architecture:
• - Backend, frontend, and
chatbot modules.
• 2. Database Design (ER Diagram):
• - User data, chat history, and
knowledge base.
• 3. Features:
• - JWT authentication, RESTful
APIs, user-friendly interface.
Deployment
• 1. Chatbot Model:
• - Hosted on InferenceAPI.
• 2. Django Backend:
• - Deployed on DigitalOcean's
Linux server.
• 3. Frontend Application:
• - Hosted on Vercel,
communicates via REST APIs.
1. Reliable chatbot
delivering credible
medical information.
2. User-friendly interface
Results for querying and
browsing health topics.
3. Improved accessibility
to healthcare
information.
• 1. Multilingual Interaction:
Enhanced global accessibility.
• 2. Voice-Based
Communication: Support for
speech queries.
Future
• 3. Offline Mode: Assistance in
Work remote areas.
• 4. Advanced Features:
Personalized health
recommendations, symptom
analysis.
• Bridges the gap in
healthcare accessibility,
• Offers reliable medical
Conclusion information.
• Future developments to
make it user-friendly and
feature-rich UI.
Ethical Considerations and
Limitations
• Purpose and Scope: • Usage Disclaimer: • Responsibility:
Basic medical information Users must consult Ensuring users do not rely on
and supports non-critical healthcare professionals for the chatbot for critical
health queries. severe or life-threatening diagnoses or treatments.
Not a substitute for conditions.
professional medical advice.
