GraphSpace

Inspiration

In an era where technology transforms lives, Alzheimer's disease reminds us of the fragility of memory and the emotional toll it takes on individuals and families. As the disease progresses, patients often lose their ability to recall cherished moments and recognize loved ones, leaving families feeling helpless and heartbroken. Our project is inspired by this struggle, aiming to provide a lifeline for those affected. By storing memory fragments, we create a digital archive that allows users to access precious moments even as memories fade. This isn't just about data storage—it's about preserving the essence of who we are and maintaining connections with loved ones. Through continuous innovation, our platform seeks to empower patients and families, offering hope and support in the face of Alzheimer's. It’s not just technology; it’s a guardian of memories, ensuring that love and connection endure beyond the grasp of this disease.

What it does

Our project is a personalized search engine that unifies your digital life by integrating data from Google Drive, Gmail, Google Contacts, and your own ad-hoc notes and ideas. Using a combination of Graph Neural Networks (GNNs) to model relationships between people, files, and topics, and Retrieval-Augmented Generation (RAG) to generate high-quality, context-aware responses, the tool offers a deeply personalized and intelligent search experience. Whether you're trying to find that one document you discussed with a colleague last year or recall a phone number from an old email, this system helps surface relevant information from across your digital ecosystem—fast, accurate, and tailored just for you.

Challenges we ran into

Accessing Gmail, Drive, and Contacts requires strict OAuth scopes and user consent. Handling sensitive data responsibly is critical. Ensuring the retriever fetches the most relevant context from personal data is hard—irrelevant chunks degrade output quality fast.

What we learned

This project taught us how powerful personal data can be when meaningfully connected. We deepened our understanding of OAuth flows and secure access to Google APIs, and learned how to preprocess and unify diverse data sources. We explored the strengths and limitations of Graph Neural Networks for modeling real-world relationships, and saw how Retrieval-Augmented Generation can produce surprisingly smart responses—when given the right context. Most importantly, we learned how to balance technical ambition with user experience and security in a short timeframe.

What's next for GraphSpace

We plan to improve the system’s scalability and optimize the graph-building pipeline to handle larger datasets efficiently. We also aim to fine-tune the retriever component for more accurate context matching in RAG. On the UI side, we’re working toward a smoother, more conversational interface. In the long run, we want to add support for more data sources (like calendars, Slack, Notion) and explore on-device or private cloud deployments to enhance data privacy and security.