SmartSummit RAG

Inspiration

Two weeks ago I visited IMEX America 2025. So many people networking and making good connections though the organizers themselves are essentially blind to all this activity. At the networking industry's premier networking event, people still wander blindly between booths, hoping for random collisions.

What it does

We implemented Document RAG for event attendees.

It sounds banal. But it pairs with event networking very effectively.

People's documents often reveal their true intent far better than any profile could. When someone uploads their strategy deck, you see:

What keeps them up at night (extracted challenges) Where they're headed (timeline extraction) Who can actually make decisions (people parsing)

How we built it

LangGraph + Temporal coordinate agentic workflows for dynamic matchmaking.

Bipartite graph-based compatibility analyses based on shared intent and complementary attributes.

Event APIs and CRMs feed data into the network.

Integrated OpenAI's Files API for multi-format support (PDF, DOCX, PPTX) Built a vector storage pipeline Created document-to-person mapping (every document reveals something about the uploader) Implemented semantic search across documents connected to user

Challenges we ran into

Figuring out how exactly to integration OpenAI's File API with our Langgraph backend to implement RAG

Accomplishments that we're proud of

Built a document ingestion pipeline that turns unstructured data into actionable and queryable information in seconds.

What we learned

Learned that document intelligence at scale creates network effects. When 1,000 attendees upload documents, each new upload makes everyone else's matches better. The graph becomes smarter.

What's next for SmartSummit

Develop “SummitGraph Analytics” — real-time metrics showing which introductions created measurable value for attendees, sponsors, and hosts.

Built With

• langgraph
• nextjs
• postgresql
• react
• temporal