🔗 GitHub
Description
mediBot is an end-to-end healthcare assistant application that provides personalized health recommendations based on patient inputs. This system utilizes a Retrieval-Augmented Generation (RAG) approach, which combines document retrieval with large language model generation. The core of the system is a medically fine-tuned LLM (Biomistral), which is tailored specifically to understand and respond to medical queries.
The application is hosted on AWS, ensuring scalability and reliability, and features an interactive UI that makes it easy for patients to interact with the system and receive real-time health advice.
Key Features
- Health Recommendations: The system uses the RAG approach to retrieve relevant medical information and generate tailored health recommendations based on user input.
- Medically Fine-Tuned LLM: The Biomistral model is fine-tuned to provide health-related insights, making it highly accurate for medical queries.
- Interactive UI: The user interface allows patients to easily input their symptoms and receive personalized advice, creating a user-friendly experience.
- AWS Hosting: The application is deployed on AWS, ensuring high availability and scalability for handling patient queries efficiently.
Technologies Used
- Retrieval-Augmented Generation (RAG) for document retrieval and LLM-based generation.
- Biomistral: A fine-tuned LLM for medical applications.
- AWS: For hosting and deployment, providing scalability and security.
- Streamlit/Flask/FastAPI: For building the interactive UI and backend services.
- Natural Language Processing (NLP) for understanding and processing medical queries.
Next Steps
- Improve the recommendation system by incorporating more health data and research.
- Expand the UI with additional features like user profile management and historical medical records.
- Integrate with medical databases for real-time health data and decision support.
- Enhance the model with continuous learning from user interactions to improve accuracy and relevance.