Node.js Developer: Real-Time Call Coaching System (Twilio + Deepgram + Claude API)

We operate a 24/7 taxi dispatch service and need a Node.js developer to build a real-time AI co-pilot system that helps our overseas call center dispatchers catch booking errors during live customer calls. THE PROBLEM Dispatchers frequently miss critical call details: wrong pickup/dropoff addresses, missed stops on multi-stop trips, wrong branch for chain businesses, and overlooked special instructions (pet transportation, wheelchair access, etc). This isn't a knowledge gap -- our dispatch software (Sherlock) has built-in Google Maps. It's an attention-to-detail problem, and real-time coaching during the call is the only point where it can actually be fixed. WHAT WE NEED BUILT An audio pipeline: Twilio Media Streams forks live call audio to Deepgram for real-time transcription, which feeds the Claude API (already have a 20-rule system prompt written) to evaluate the call against best practices and push correction "script cards" to a permanent always-on browser window on each dispatcher's workstation in real time. A secondary verification layer: our dispatch software (Sherlock/Postgres) fires a LISTEN/NOTIFY event the moment a booking is created. Your code needs to subscribe to that event, compare the completed booking against the call transcript, and flag mismatches before a driver is dispatched. WHAT'S ALREADY DONE - Full technical architecture and 15-page developer spec (attached) - Complete 20-rule system prompt for Claude - Working React prototype for the dispatcher-facing UI (transcript panel + script card panel) - Twilio account active with valid credentials - Sherlock/Postgres integration path confirmed by their PM, handled separately from your quote - Sample booking data schema from our dispatch software (attached) WHAT YOU'LL DELIVER - Backend Node.js app connecting Twilio, Deepgram, and Claude APIs via WebSocket - Postgres LISTEN/NOTIFY subscriber and mismatch-detection logic - Integration of the provided React UI with your backend - Docker setup and deployment documentation for our VPS - Basic monitoring/logging TIMELINE & BUDGET Expecting about 4 weeks from kickoff to production. Budget above is a starting estimate -- please send your own fixed-price quote with your proposal. TO APPLY Please review the attached RFP package (full architecture, data schema, rulebook, and a proposal template) and respond covering: your relevant experience (especially Twilio Media Streams, WebSockets, and Postgres LISTEN/NOTIFY if you have it), your technical approach, timeline, fixed-price quote, and 2-3 portfolio links or references.