Project Title: ARIA NEXUS: A Fully Offline Bilingual Voice AI Assistant on Edge Hardware
Description:
ARIA NEXUS is a fully offline, bilingual (Arabic and English) voice AI assistant that runs entirely on a single NVIDIA Jetson Orin Nano 8 GB edge device, with no cloud connection required during operation. It integrates seven local services—wake-word detection, Whisper speech recognition, a quantized Qwen2.5 language model served through llama.cpp, Piper speech synthesis, Chroma retrieval-augmented generation, a smart-home controller, and a touchscreen dashboard—together with an audio-reactive animated face, persistent personal memory, user-defined routines, and barge-in interruption. By performing all recognition, reasoning, retrieval, and synthesis on-device, the system removes the privacy exposure, latency, and connectivity dependence of commercial cloud assistants such as Alexa, Siri, and Google Assistant. It responds to spoken commands in one to four seconds and switches automatically between Arabic and English. Built on hardware costing approximately 450 US dollars, ARIA NEXUS demonstrates that a practical, privacy-preserving conversational assistant can run on affordable, low-power edge hardware. This paper presents the problem, the proposed solution, the methodology, and the demonstrated results, reporting reliable offline operation, 98 percent wake-word accuracy, and working bilingual, memory, and smart-home functionality.
Uploaded At: 2026-07-05 04:15:28