Prototype to Production: Building Enterprise MCP and AI Agents with Templates - Tuhin Sharma

This talk discusses the failure of over 95 percent of GenAI pilots to reach production, highlighting the importance of engineering discipline in overcoming this challenge. The speaker presents a practical blueprint for building production-ready Model Context Protocol (MCP) servers and AI agents using two refined open-source templates based on real-world enterprise deployments. Attendees can expect detailed code walkthroughs and live demonstrations that cover FastAPI-based MCP server architecture, streaming agent implementations with PostgreSQL, and observability through Langfuse tracing. The session also addresses Kubernetes deployment patterns, rootless container configurations, and automated recovery strategies, equipping participants with production-grade templates and engineering patterns that can transform prototypes into reliable enterprise systems.