Why LinkedIn Treats AI Agents as Infrastructure - Not Features

Explore LinkedIn's strategic approach to AI agent implementation in this InfoQ podcast episode featuring Karthik Ramgopal and Prince Valluri discussing how the company treats AI agents as infrastructure rather than standalone features. Learn how LinkedIn transitioned from isolated proof-of-concepts and "hermit developer" approaches to building a unified Platform Engineering framework for multi-agentic systems. Discover the Model Context Protocol (MCP) and understand the critical distinction between "foreground" IDE-based agents and "background" asynchronous agents that operate independently. Examine how specifying intent through specifications and sandboxes creates a structured approach to AI-driven software delivery, while exploring the essential "human-in-the-loop" review processes that ensure quality and compliance. Gain insights into implementing security, compliance, and observability measures for AI systems at enterprise scale, and understand how LinkedIn leverages RAG (Retrieval-Augmented Generation) and historical data to solve context challenges. Access practical advice for Platform Engineers looking to build secure, scalable AI infrastructure that moves beyond individual developer scripts to enterprise-ready solutions.