Work

Voice ordering AI for kiosks

A full-duplex, real-time voice agent that takes restaurant orders at unattended kiosks.

Role
Senior software engineer, device team
Stack
Flutter, LangChain, Deepgram, native audio
Year
2025 to now
Status
In production

At 365 Retail Markets I took a real-time voice ordering agent from proof of concept to production. The agent grounds its ordering in menu data through a RAG pipeline, streams speech to text and text to speech, and keeps per-session context and cost controls.

The hard part was audio. I built the native audio plugin with acoustic echo cancellation and mid-speech interruption, so a customer can talk over the agent and be heard. No screenshots here, it is employer work; the description matches my public CV.