An AI prototype that looks remarkable in a controlled demo can struggle for months to make it into production. The real challenge begins when that prototype has to operate within the constraints of enterprise systems, data, security, and governance.
This session explores four core pillars of production-grade AI systems: context engineering, evaluations, memory, and security and governance. We'll examine the common pitfalls teams encounter, why they matter, and the practices that make AI systems reliable in production.
Designed for engineering teams building beyond prototypes, this talk provides a practical framework for creating AI systems that are reliable, measurable, secure, and ready for real-world use at scale.
Shubham Jindal is Director of AI Engineering at Harness, where he leads AI initiatives focused on enterprise AI systems and agent-based architectures. Previously, he was a Founding Engineer at Traceable AI and spent nearly three years at AppDynamics working on browser monitoring and web performance technologies. An IIT Delhi Computer Science graduate, Shubham is also an inventor with multiple patents in AI and anomaly detection, and a frequent speaker on AI agents, MCPs, and modern software systems.
