HeartLink

Inspiration

We were inspired by thinking about how to incorporate an LLM into a healthcare hack and recognized that dementia patients in long-term care facilities are a group in need, often without the level of care they need due to lack of resources. We then decided to develop an application to provide personalized support through an LLM chatbot tailored to their memories and experiences in order to enhance the quality of care for these patients.

What it does

The HeartLink website allows family members to log in and upload memories as text that would be helpful for aiding their loved one's memory who is living in a long-term memory care facility. The caregivers of the patients would then also be able to log into HeartLink where the patient can chat with Link, the chatbot, who tailors the conversation specifically to the patient's memories that their family input.

How we built it

We used Next.js application for the UI and frontend, and used AWS Bedrock to build the RAG LLM chatbot and for database storage.

Challenges we ran into

We ran into issues connecting the RAG LLM to the data storage properly and also getting the LLM output to the chatbot screen.

Accomplishments that we're proud of

Functioning UI, user authentication, as well as a cohesive UI design. We are also proud of the chatbot interface where Link's face "speaks".

What we learned

How to build a RAG LLM, build a frontend using Next.js and React, and upload data to database.

What's next for HeartLink

Add audio to text functionality enabling users to more easily interact with the chatbot instead of typing and implement photo upload and chatbot display functionality to increase memory support for patients.

Built With

• amazon-web-services
• aws-bedrock
• jsx
• next.js
• react.js
• titan-text-embedding-v2
• typescript