An AI-native stack for places people walk into.
Avyra is a separate company building seven services that share one brain — vision sees the guest, the brain decides, the avatar speaks, the room bridge listens, the voice agent answers the phone, the POS rings it up, and the dashboard tells the operator what just happened. Poised Genie is the engineering team behind it. The pages that follow describe what we shipped.
One contract. Many surfaces.
Every Avyra service publishes and subscribes through a single Redis-backed event bus. avyra:presence:{tenant} is who's in the room. avyra:utterance:{tenant} is what gets said. avyra:speech_state:{tenant} keeps the mic and the speaker from talking over each other. No hardcoded copy lives in any room device — every word comes from the brain.
- Vision publishes presence events from a USB camera.
- Room-bridge mediates mic, VAD, STT, and turn-taking.
- AI-engine decides what to say and when to transfer.
- Avatar speaks with lip-synced wave rings.
- Voice AI handles the phone with the same brain.
- Dine + POS settle the actual transaction.
Guest walks in
│
▼
[ avyra-vision ] ─ publish presence ─► Redis
│
┌───── subscribe ───────┤
▼ ▼
[ avyra-room-bridge ] [ avyra-avatar ]
│ ▲
▼ │
┌──── /api/room/turn ──── [ ai-engine ]
│ │
▼ ▼
[ STT :8001 ] [ TTS :8002 ]
│
▼
Speaker / phone
Phone calls follow the same path,
just terminated by Asterisk + PJSIP.Pick one. Or run the whole suite.
Each Avyra product ships as its own service with its own contract. Adopt one for an immediate win, or compose the full room when you're ready.
Got a product to build? Tell us what you have in mind.
We kick off in days, not months. Working software in weeks. If we're not the right fit, we'll tell you up front.