VOX — Multilingual Voice & Document AI Assistant

VOX (ASRRS) is my final-year project: a multilingual admissions assistant working in English, Urdu and Roman Urdu. It combines intent routing, document retrieval, local language models and a low-confidence human handoff.

The implementation uses speech recognition, Ollama, ChromaDB, Asterisk/SIP and text-to-speech components. The query pipeline checks intent and scope before retrieval, then uses source material to support an answer.

My work covers the conversational pipeline, retrieval, local deployment and telephony integration. The system is a project prototype; answer quality and the phone flow need evaluation for each new deployment.

The visual is a technical overview, not a live product screenshot. No private university documents, caller information or credentials are shown.