At ClaudiaMind, I architected a voice-based RAG system that answers phone calls by retrieving context from business knowledge bases and generating grounded spoken responses in real time.
I optimized the voice assistant by benchmarking speech-to-text, language model, and text-to-speech providers, selecting combinations to achieve sub-two-second response latency while minimizing cost per conversation.
I designed and deployed a Python voice-agent backend for the WhatsApp Business Platform, using a WebRTC server for real-time audio processing. I also built interactive message flows, command handlers, and automated survey and queue management with in-chat metrics reporting.
At Cheetrack, I built and maintained messaging features for a WhatsApp-based sales and customer-service platform using Twilio, Python, and React. Earlier, I developed a Bitcoin purchase-management bot at ProIt and migrated Leadbox's car-dealership inventory system from AngularJS to Vue.js.
