AI Calling Agent - Autonomous Security Copilot

Aegis Nexus AI is an autonomous, real-time incident response voice agent for enterprise SOCs, deployed on Vultr cloud infrastructure. THE PROBLEM: Enterprise server downtime costs roughly $5,600 per minute. When zero-day breaches occur off-hours, SOCs rely on passive alerts (email/Slack). By the time an engineer reads the alert, investigates, and deploys a patch, massive financial damage has occurred. THE TECHNICAL SOLUTION: Aegis bridges AI threat detection and human execution. Instead of emailing, our Node.js backend triggers a Twilio voice call to the on-call engineer. To achieve human-like latency, we bypassed Twilio's standard TwiML. We architected a continuous WebSocket bridge that pipes raw 8kHz mulaw audio directly into the Speechmatics Real-Time API. As the engineer speaks, transcriptions route to Google Gemini 2.5 Flash. Using native function calling, Gemini can: - search_threat_intel: Query the SQLite knowledge base - get_active_incidents: Check real-time server status - send_incident_report_whatsapp: Dispatch crash logs to WhatsApp - escalate_to_level3: Transfer to a human manager Responses are synthesized via ElevenLabs' Multilingual V2 model, automatically adapting to English, Hindi, or Marathi. VULTR DEPLOYMENT: The entire architecture—Express.js server, WebSocket streamers, and React dashboard—is fully deployed on a Vultr Compute instance using SQLite WAL. Aegis Nexus AI reduces critical incident response time from 30+ minutes to under 60 seconds.