AAP.JS — Voice Interaction System

Senior Design capstone: an embedded voice interaction system running on a Raspberry Pi 4, designed for accessible, hands-free conversation in public/expo settings.

Speech is captured and transcribed with Azure Cognitive Services Speech-to-Text, benchmarked at 97.94% word accuracy on field recordings. Transcripts are routed to the DeepSeek API for context-aware reasoning, with a lightweight intent-routing layer to keep latency low.

Sustained 3+ hours of continuous, stable public interaction at the Senior Design Expo with no manual restarts — proving robustness of the audio pipeline, error recovery, and prompt-state management.