Repaire
Team consisting of Hugo Carpentier, École 42 student (coding-focused), and Maxime Biraud, ESCP Business Analytics & AI student and ex-Richemont/Cartier luxury-tech product innovator.
Loom Video
Project Description
“Scheduling should be as simple as talking. Meet KAL—the talk-to-schedule assistant. You speak; it books, moves, summarizes, and fixes problems—end-to-end.”
“Managing one or multiple calendars eats time. Links and forms break the moment you need negotiation, travel time, or last-minute changes. You end up organizing your life instead of living it.”
“With KAL, you press Talk and say: ‘Schedule 45 minutes with Hugo this afternoon.’
KAL checks your calendars, proposes two clean options, confirms by voice, creates the Meet link, and sends invites.
Anytime you can ask: ‘What’s my next three days?’ and KAL returns a crisp three-bullet summary.”
What’s new
“Three things make KAL different:
Voice-to-voice loop—a real conversation that actually closes the booking.
Proactive intelligence—live routes, ‘leave now’ nudges, conflict detection, instant re-booking when plans change.
Memory—with Qdrant we learn your preferred times, busy periods, frequent attendees, and typical durations to deliver a hyper-personalized experience.”
“On Google Cloud, we use streaming Speech-to-Text; ElevenLabs for natural TTS; an orchestrator talks to Google Calendar.
Qdrant holds two collections:
conversations: transcript, intent, duration, confidence, timestamp, 768-dim embedding;
user_preferences: preferred meeting windows, common attendees, typical durations, scheduling habits.
The UI is a minimal black-and-white clock with a single Talk button.”
“In this sprint, we focused on the core experience and architecture. Some modules—like full multi-party negotiation—are still in progress. But the direction is clear.”
“If you want a calendar that thinks ahead and speaks your language: Speak. It’s scheduled. We’re looking for feedback and early users.”