Confer

AI-driven platform for multi-day conference orchestration.

Comment

Inspiration

Organizing a multi-day conference is a logistical nightmare involving countless moving parts: finding sponsors, curating speakers, securing venues, managing ticketing, determining pricing, and executing go-to-market strategies. I realized that a multi-agent AI system could perfectly model these distinct roles, operating autonomously yet collaboratively to plan an entire event from scratch.

What it does

Confer is a 7-Agent Conference Intelligence platform that automates the full lifecycle of organizing events. An Orchestrator agent coordinates 6 specialist agents to handle the workload:

  • Sponsor Agent: Identifies and drafts sponsor outreach.
  • Speaker Agent: Curates speakers based on topic and audience.
  • Venue Agent: Searches and ranks venues.
  • Ticketing & Pricing Agents: Manages tiers, allocations, and dynamic ROI modeling.
  • GTM & Ops Agents: Builds marketing campaigns and manages live logistics.

How I built it

I architected the backend using FastAPI to power the 7 autonomous agents. The agents leverage Groq for fast LLM inference and Google Gemini for complex reasoning. For external tools, I integrated Tavily for web search and the Google Places API for venue discovery.

The system relies on Supabase for persistence and authentication, alongside ChromaDB as the vector store. The frontend is a modern web app built with React 18, Vite, TypeScript, and TailwindCSS, utilizing Zustand for state management and Recharts for analytics. The frontend and backend communicate via WebSocket streaming to provide users with real-time updates as the agents work and collaborate.

Challenges I ran into

Coordinating 7 autonomous agents in real time was a significant challenge. Ensuring the Orchestrator agent effectively routed tasks and aggregated outputs without creating bottlenecks or infinite loops required careful prompt engineering and state management. Additionally, managing WebSocket connections for real-time frontend updates proved challenging when dealing with multiple concurrent agent actions.

Accomplishments that I'm proud of

I'm incredibly proud of the seamless integration of the multi-agent system. Watching the agents collaborate—for example, the Venue Agent finding a location which then informs the Pricing Agent's ROI model—is deeply satisfying. I'm also proud of the real-time dashboard that visualizes this orchestration beautifully.

What I learned

I learned a great deal about designing multi-agent architectures, particularly how to define clear boundaries and responsibilities for each agent using the specialist model. I also deepened my knowledge of real-time web technologies (WebSockets) and state management in complex React applications.

What's next for Confer

I plan to integrate more specialized agents (such as a Catering Agent and a Travel Agent), add support for managing virtual and hybrid events, and integrate directly with CRMs and email marketing platforms so the system can actively execute the campaigns designed by the GTM agent.

Built With

Share this project:

Updates