Voice AI agent for real estate inbound calls — lead qualification, scoring, and CRM sync
Real estate companies had no intelligent voice agent that could handle inbound calls, qualify leads, and deliver actionable analytics. Every client uses different CRMs, call flows, and scoring criteria, so the system had to be built as a productized service with a fully modular codebase that can be configured per client without rewriting anything.
Solely owned frontend, backend, DevOps, and GenAI integration from scratch with no handoffs. Built on LiveKit for real time voice, Twilio for phone number provisioning and inbound call routing, Auth0 for role based access, and HubSpot for CRM integration.
Architected the system for configurable per client deployment so new real estate companies could be onboarded from a single codebase without rebuilding core infrastructure. Each client got their own configuration for call flows, lead scoring rules, and CRM sync.
Built persistent caller profiles that capture phone number, name, intent, and key real estate signals across calls to automatically score leads and give companies a clear signal on who to prioritize.
Implemented regex based STT validation to ensure transcription accuracy before passing audio input downstream, and built real time knowledge base highlighting so the agent surfaces relevant context during live calls.
The LLM powered both real time conversational responses during calls and post call analysis for lead scoring and conversation level insights.
Shipped a production prototype used directly in client discovery sessions. Built an admin system where admins can generate shareable demo links giving clients a guided view of the full dashboard, plus an analytics dashboard for call performance and conversation insights.