Voice-first AI front desk for a hotel
A lobby kiosk and a phone receptionist that handle check-in, walk-ins, stay extensions, checkout and room changes in 5 languages.
- Role
- Built the kiosk app, backend, PMS microservice, ID OCR and RAG concierge
- Scale
- 5 languages, at the lobby kiosk and on the phone
The problem
A hotel front desk spends its time on a handful of routine requests: check-in, walk-ins, extending a stay, checkout and room changes. They come at the desk and over the phone, and in the guest's own language.
What I built
- The kiosk single-page app and its backend.
- The PMS microservice.
- Passport and ID OCR with OpenAI vision and MRZ parsing.
- A RAG concierge over Qdrant.
Technical details
- Backend in Rust with Axum on PostgreSQL; kiosk in React 19.
- Voice through OpenAI Realtime, Gemini Live and LiveKit.
- Integrations with Stripe, Twilio and WhatsApp.
- Concierge answers retrieved from Qdrant, at 93% hit@1.
Stack
- Rust
- Axum
- PostgreSQL
- React 19
- Qdrant
- OpenAI Realtime
- Gemini Live
- LiveKit
- Stripe
- Twilio