From Raw Audio to Real-Time Insights: Building a Gen AI Call Analysis Pipeline

This demo showcases how to build a complete gen AI-powered pipeline for analyzing customer calls—both in batch and real-time—using MLRun. Without writing custom code, you’ll see how to:

• Automatically transcribe calls, detect PII, and anonymize sensitive data
• Classify call topics, summarize conversations, and evaluate sentiment or agent performance
• Put together reusable, open-source functions from MLRun’s Function Hub to rapidly build your pipeline
• Query structured call metadata through a front-end app and play real audio alongside model insights
• Transcribe live audio in real time using Whisper, passing text directly to downstream LLM-powered apps for co-pilot use cases

Whether you’re building offline analysis tools or responsive call center co-pilots, this demo offers a practical look at how to go from unstructured audio to actionable insights—fast.