AI Document Intelligence & Voice Assistant
Natural-language search across compliance documents, paired with automated outbound voice reminders.
Project Overview
Compliance teams needed answers out of long PDFs without reading them end to end. A RAG pipeline built on Amazon Bedrock with OpenAI vector storage answers natural-language questions across multi-page documents, and a voice layer using Vapi AI webhooks with Node-cron scheduling places outbound appointment confirmation calls automatically.
Quick Facts
Answers buried in multi-page compliance documents
Compliance material sat in long PDFs that had to be read manually, while appointment confirmations depended on someone remembering to call.
RAG retrieval plus scheduled voice automation
Documents are embedded into vector storage and queried through a Bedrock and OpenAI RAG pipeline, returning answers grounded in the source text. Vapi AI webhooks driven by Node-cron handle outbound appointment confirmations on schedule.
