AI call assistant

Year

2025

Tech & Technique

Python, Twilio, Amazon Bedrock, ElevenLabs TTS, AWS Transcribe, LangGraph, Langfuse, PipeCat, OpenSearch, Google Calendar API

Description

An AI-powered voice assistant built for home sales in Florida, designed to handle inbound calls and guide prospects through a human-like conversation to book community visits. Powered by Twilio for telephony, Bedrock-hosted LLMs for dialogue, and ElevenLabs for ultra-realistic voice output. Integrated with OpenSearch for knowledge retrieval and Google Calendar for appointment booking.

Key Features:

📞 AI Voice Assistant: Handles inbound calls with natural, conversational tone
🗣️ Realistic TTS/STT: ElevenLabs for expressive speech and AWS Transcribe for accurate recognition
🧠 Smart Dialogue Control: LangGraph-powered workflows for guided, dynamic conversations
📅 Calendar Booking: Real-time event creation with Google Calendar and SMS confirmations
🏘️ Multi-Community Support: Pulls structured info from different home communities using RAG
📊 Observability: Integrated with Langfuse for prompt tracking, metrics, and debugging

Technical Highlights:

Used Amazon Bedrock-hosted LLM for consistent and cost-effective language generation
Leveraged Twilio Programmable Voice for call routing and audio streaming
Built Python backend to manage conversation flow, STT/TTS, and calendar events
Integrated OpenSearch as a vector-based knowledge base for real-time context grounding
Employed LangGraph for multi-turn state machine control and fallback handling
Used PipeCat for real-time audio stream handling between Twilio and ElevenLabs

My Role

Full-Stack AI Engineer
Led the full lifecycle of voice assistant development and deployment:

✅ Backend: Developed a Python server for managing voice inputs, LLM prompts, and event creation
📞 Voice Pipeline: Integrated Twilio, ElevenLabs TTS, and AWS Transcribe for fluid call experience
🧠 LLM Integration: Used Amazon Bedrock with LangGraph and Langfuse to control conversation logic
🔍 RAG System: Connected to OpenSearch vector DB for real-time community info retrieval
📅 Calendar Integration: Synced appointment bookings directly with Google Calendar + SMS confirmation
🚀 Deployment: Hosted services using AWS infrastructure with observability and error tracking