Inspiration
Dementia and Alzheimer’s don't just erase memories; they erase independence. We were inspired to help families by creating an AI companion that doesn't "forget" after every session, providing a persistent digital memory for those who need it most.
What it does
CognitiveCare is an AI agent that remembers a patient’s life facts, family, and preferences. It tracks medication adherence, monitors therapeutic goals through natural conversation, and provides a dashboard for doctors to manage care and track progress.
How We Used MemVerge&Neo4J
We utilized MemMachine from MemVerge for persistent dual-layer memory and Neo4j Graph DB (Aura DB) to map complex relationships between patient history and daily interactions. The backend is powered by FastAPI, the UI by Streamlit, and the entire stack is deployed on AWS EC2 instances.
Challenges we ran into
The biggest hurdle was Smart Routing—ensuring the AI correctly distinguished between temporary daily logs and permanent profile facts. We also focused on optimizing the connection between the Neo4j knowledge graph and the MemMachine layer to minimize latency.
Accomplishments that we're proud of
We successfully built a "Memory That Lasts." Seeing the AI naturally reference a patient’s daughter or a specific hobby in a helpful, empathetic way—backed by a robust graph database architecture—felt like a genuine breakthrough.
What we learned
We learned the critical importance of persistent state and relationship mapping in healthcare AI. We also gained deep experience in deploying containerized applications on AWS and managing the integration of high-performance memory tools with Aura DB.
What's next for CognitiveCare
We plan to implement Voice Interface for better accessibility, Advanced Analytics to detect early cognitive decline from speech patterns, and more granular date-based medication scheduling for doctors.
Built With
- amazon-web-services
- ec-2
- gemini
- memverge
- neo4j
Log in or sign up for Devpost to join the conversation.