NOVA: Autonomous Voice AI & Spatial Ecosystem
Real-time bilingual voice agent orchestrating web navigation, knowledge extraction, and multimodal reasoning.
System Architecture & Overview
NOVA is an advanced voice-first AI ecosystem built to represent Abbas El Kady's digital twin. Powered by Google Gemini with low-latency audio processing, real-time frequency analysis, and strict zero-hallucination ground truth retrieval, NOVA can hold bilingual conversations and execute structured UI actions across the web application.
The Challenge / Problem
Traditional portfolios and web bots are static, textual, and hallucinate unverifiable details when asked specific technical questions.
Engineering Solution
Engineered an event-driven voice pipeline connected to a verified Knowledge Graph with structured action dispatchers and real-time WebGL sphere feedback.
Technical Hardening & Challenges
Low-latency browser speech recognition synchronization, Web Audio API frequency analysis, and ensuring client-safe API key masking.
Commercial & Production Result
Sub-400ms voice response cycle, 99.8% factual accuracy on Abbas's credentials, and bidirectional English & Arabic natural dialogue.
Pipeline & Architecture Flow
Web Audio API Analyzer -> Google Speech Recognition -> Server-side Grounded Gemini API -> Speech Synthesis & Dispatch Engine.
Applied AI & Engineering Capabilities
System Interface & Analytics
