AI token maxing - Paris Vibe-a-thon: “Build a One-Person Unicorn” — November 15, 2025
AI Tinkerers - Paris
Hackathon Showcase

AI token maxing

Team consisting of AI engineers Gaspard Hassenforder (Veltys) and Alex Bohane (Owkin), HEC/École Polytechnique MScs—Python, LangChain, RAG, Neo4j, agentic/RL, BigQuery.

2 members Watch Demo

Project Description

Our project is an AI Event Planner that organizes real-world events end-to-end from a single conversation. Event organizers—community builders, startup ops teams, HR departments, and coworking spaces—simply describe what they want once (“Plan a 24-hour AI hackathon for 80 people in Berlin with a €5,000 budget”), and our system autonomously generates a complete, actionable event plan including: venue options with pricing, event schedule, sponsor outreach materials, participant targeting strategy, promotional posters, and even makes live phone calls to venues to check availability and negotiate pricing.

Core Problem:

Event planning today is painfully manual. Organizing even a small meetup or hackathon requires juggling spreadsheets, endless email threads, phone tag with venues, sponsor outreach, attendee management, and marketing assets—often taking 20-40 hours of scattered work. Most small events never happen because the operational burden is too high, and existing tools (Eventbrite, Notion, etc.) are coordination platforms, not intelligent agents that do the work for you.

Target Customer:

  • Tech community organizers running recurring meetups/hackathons

  • Startup operations teams organizing offsites and internal events

  • Coworking spaces offering “event-as-a-service” to members

  • HR/People teams at scale-ups running quarterly team-building events

  • Event agencies looking to automate repetitive planning workflows

Main Features:

  1. Conversational Event Briefing: Natural language input via chat (extendable to voice)
  1. Orchestrator Agent (Mistral): Breaks down the event into structured tasks: venue research, sponsor targeting, participant acquisition, marketing assets, and outreach calling
  1. Research Agent (Mistral): Analyzes the brief and generates venue shortlists, sponsor segment recommendations, and participant targeting strategies
  1. Visual Agent (fal.ai): Automatically generates promotional posters and event branding assets
  1. Voice Agent (ElevenLabs): Places autonomous phone calls to venues to inquire about availability, pricing, and negotiate rates if over budget
  1. Structured Output: Returns a complete event blueprint (JSON + human-readable) with schedule, budget breakdown, outreach scripts, and booking next steps

Demo Flow (what judges will see):

  1. User inputs: “Plan a 24-hour AI hackathon in Berlin for 80 people, budget €5k, focus on fun and memes”
  1. Orchestrator Agent (Mistral) analyzes the request and decides: need venue research → sponsor targeting → participant strategy → promotional asset → venue calling
  1. Research Agent (Mistral) returns: 4 candidate venues in Berlin (co-working spaces, tech hubs), 3 sponsor segments (AI companies, dev tools, cloud providers), and participant personas (AI engineers, startup founders, students)
  1. Visual Agent (fal.ai) generates a vibrant, meme-friendly hackathon poster with event branding
  1. Voice Agent (ElevenLabs) makes a live call to the top venue: “Hi, I’m calling on behalf of an AI hackathon organizer. Do you have availability on June 15th for 80 people? What’s your rate?” → negotiates if too expensive → confirms availability
  1. System assembles everything into a clean event plan with: venue confirmation, sponsor outreach emails, participant recruitment strategy, visual assets, and a spoken summary of next steps

built off a repo that was using chat kit for our front end

ElevenLabs Fal Mistral AI