BrainClone | Devpost

Inspiration

We forget opportunities, connections, and moments that could matter more. We often get overwhelmed and lose focus, not because we’re unproductive, but because our brains are overloaded. Memory is powerful, but human recall is limited in today's fast-paced world full of notifications, new news, new ideas, thoughts, and noise. Our minds are overflowing. On top of that, under stress, memory recall has proven to weaken by research.

So we asked, What if your brain had help remembering what matters? What if your experiences could be captured, organized, and retrieved dynamically from an AI-powered second brain?

What it does

You give the input. Journal freely about your ideas, memories, and emotions. We capture your thoughts and understand context using AI. It extracts meaning, connections, and insights. We build your second brain. Every memory becomes a node in your personal network. You find clarity. Ask questions like “When did I feel most inspired?” and get answers from your own past. We retrieve related data from your digitally stored memory and present it to you, with answers.

How we built it

Frontend: Next.js, React, TailwindCSS, React-Force-Graph-3D
Backend: Opik, FastAPI (Python 3.13), Structlog
Data: Neo4j (relationships), PostgreSQL + pgvector (embeddings), Redis (cache)
AI Stack: R2R for text ingestion, NER for understanding entities, Embeddings + RAG for semantic memory recall, Gemini API

key endpoints

GET /health - service health
GET /metrics - vector and graph counts
POST /api/v1/documents/upload - upload and process document via r2r
GET /api/v1/documents/{id} - document metadata
POST /api/v1/documents/create-graph - build graph from documents
POST /api/v1/graph/entities - create entity
GET /api/v1/graph/entities/{id} - fetch entity and relationships
POST /api/v1/search/hybrid - hybrid search across documents and graph

Challenges we ran into

We discovered that the 3d-force-graph library is not deployable on React Native, which means we’ll keep mobile deployment for the future. For now, we are hosting the project as a web application. Another major challenge was our brainstorming process. It was long and sometimes difficult, since each team member brought very different perspectives and opinions. Finding alignment took hours, but it was ultimately worth it.

Accomplishments that we're proud of

We successfully graphed memories, people, events, and locations into a space-like 3D knowledge graph.
We built a RAG pipeline capable of answering questions about a person’s life based on graph data.
We deployed four individual microservices, all interacting with each other through APIs.
We learned and applied new tools under pressure, including Neo4j, which at first felt intimidating.
Our front-end design turned out exactly as intended, delivering the experience we envisioned.
Despite brainstorming challenges, we found middle ground, collaborated effectively, and delivered an MVP we’re proud of.

What we learned

Keeping the team environment light-hearted, humorous, and unhinged was the best way to resolve conflicts smoothly.
For hackathons, lightweight deployments (like ngrok) are often more efficient than spinning up full-scale AWS instances (e.g., EC2).
We practiced deploying a complex system that included Neo4j, PostgreSQL, react-force-graph, and Next.js, gaining valuable hands-on experience.

What's next for BrainClone

Mobile deployment for easier access and wider adoption. We would like to send users Smart Notifications about following up with people, improving networking and productivity, and prompting users for a deeper reflection when they enter the app.