This project was created by community members Mostafa Alazazy and Augusto Gonzalez Bonorino.

Problem

Healthcare can be an overwhelming and time-consuming process for patients, leading to longer wait times, paperwork, and ultimately higher costs. Moreover, there is a need for virtual assistants in developing countries to supplement the lack of labor supply or poor level of education.

Solution

Health-E is an AI healthcare assistant that aims to cut down healthcare costs by reducing queue times, paperwork, and increasing efficiency overall. It combines a custom conversant persona and grounded QA to aid in processing patients, extracting relevant information to fill out forms, and providing advice on applicable scenarios if prompted with a question. This technology can be extended beyond healthcare to benefit regional economies and local communities in developing countries that have wide internet access.

Tech Stack

To build Health-E, the team used the following tech stack:

• google-search-results: A Python library that allows the scraping of Google search results.
• bs4: A Python library that makes it easy to scrape information from HTML and XML documents.
• pydantic: A data validation and settings management library that provides runtime checking and generation of documentation.
• streamlit: A Python library that allows for the creation of interactive web apps for data science.
• streamlit-chat: A Python library that allows for chatbot functionality in Streamlit.
• pandas: A Python library that provides data manipulation and analysis tools.
• numpy: A Python library that provides numerical computing tools.
• pillow: A Python library for handling images and image processing.
• typing: A Python library that provides support for type hints in Python.
• cohere: An API that provides AI language models to build conversational interfaces.

Cohere's Endpoints

Cohere's API was used to create Health-E's conversational interface. The following Cohere endpoints were used:

• PromptChatbot: Health-E is essentially a custom persona armed with google search. We leveraged Cohere’s PromptChatbot class, which implements the Generate endpoint, to design persona that acts as a nurse assistant capable of extracting relevant notes from conversations with patients.
• GroundedQA: Patients may seek short-term advice on what to do before their appointment. To implement this functionality Helath-E leverages an simple question recognition logic with regex and Cohere’s GroundedQA bot, which implements both the Generate and Embed endpoints, to generate a referenced response to a question.

Inspiration

The Health-E AI assistant was inspired by the need to improve the healthcare experience for patients and reduce costs. The team also envisioned using this technology in developing countries to supplement the lack of labor supply or poor level of education. The goal of the project is to make healthcare more accessible and efficient for everyone.