Talk-Lee: AI Voice Agent That Answers Business Calls

Every business phone line has the same weak spot: calls come in when nobody can pick up. After hours, at lunch, during a rush. Each one is a missed booking, a support question left hanging, or a lead that goes cold before anyone calls back.

Talk-Lee is an AI voice agent platform that answers and makes business calls for scheduling, support and lead qualification. As Lead AI Engineer, I built the real-time voice pipeline in Python.

Streaming speech-to-text, so the agent hears the caller while they are still talking

An LLM and NLU layer that works out intent and keeps context across the whole call

Natural text-to-speech, so replies sound like a person and not a phone menu

A call orchestrator for routing and clean handoff to a human when the call needs one

Integrations with Twilio, HubSpot and Calendly, so bookings and leads land where the team already works

Under 500ms response time

1,000+ concurrent calls

30+ languages

GDPR and TCPA compliant

Delivered white label for healthcare, real estate and finance

Voice AI lives or dies on latency. Callers forgive a lot, but not silence. The model was the easy part. The real work was making every step stream, from speech in to speech out, so the agent answers as fast as a person would.

If your team is losing calls, or paying people to answer the same five questions all day, this is the kind of system I build.