A voice AI agent that autonomously makes phone calls to gather customer feedback — conducting natural conversations, transcribing responses, and storing structured data for analysis. ● Built on Pipecat (voice AI pipeline framework) integrated with Twilio for outbound phone call management — handling call initiation, audio streaming, and session lifecycle. ● Engineered real-time STT/TTS pipeline enabling natural back-and-forth conversation flow with customers, with dynamic prompt injection based on feedback context. ● Stored conversation flags, full transcripts, and audio recordings in MongoDB with audio files archived to AWS S3 for durable, queryable storage. ● Resolved breaking API changes in Pipecat after a version upgrade under time pressure — debugging against library source code and Claude-assisted function reconstruction. ● Designed for business use cases: post-purchase feedback, NPS surveys, service quality assessments. Tech stack: Python, Pipecat, Twilio, MongoDB, AWS S3, OpenAI (STT/TTS), Node.js