Restaurant techOn-Device Restaurant Assistant
A Restaurant iPad That Runs Its Own AI, Offline
An iPad demo that runs Llama 3.2 1B and Whisper on the device for order-taking, a menu kiosk and end-of-shift summaries.
- SpeakVoice order on the iPad
- TranscribeWhisper Tiny on device
- ReasonLlama 3.2 1B with the menu in its prompt
- ShowOrder, upsells and allergen banner
Rochester, N.Y. A point of sale can't stop working when the Wi-Fi drops, so I built a demo where the AI runs on the iPad itself. It has three tabs. The server assistant parses natural-language orders, suggests upsells and raises allergen alerts. The customer kiosk answers menu questions. The manager tab summarizes the shift.
Llama 3.2 1B and Whisper Tiny run through react-native-executorch. Each mode has its own system prompt with the full sample menu (about 30 items with allergens, modifiers and pairings) built in, plus a rolling 6-message history. Every screen shows an offline badge and live tokens per second, so someone non-technical can see the model really is local. It was a one-day pitch demo, so the menu and the shift data are samples. The first launch downloads roughly 1.5 GB of model weights.