Mythili Annamalai Sekar

Title of the Talk: TBD

Abstract :
AI systems are no longer experiments, they are making decisions in enterprise production environments every day. But enterprise AI agents fail in production, not because the models are weak, but because the systems around them remain invisible to the teams operating them.​ Traditional monitoring tells you a system is running. It does not tell you why a retrieval pipeline returned stale context, why an agent called the wrong tool in sequence, or why model behavior quietly drifted weeks after deployment with no alert ever fired. These are not edge cases. They are documented, recurring failure patterns and most organizations have no systematic way to catch them.

AI observability is the discipline that closes this gap. This keynote examines what it truly means to make AI behavior legible in production: end-to-end tracing that reveals what logs cannot, retrieval quality signals that surface failures before they reach users, behavioral baselines that make drift measurable, and continuous evaluation pipelines that replace the dangerous assumption that a system that passed pre-launch testing will stay reliable. In enterprise AI, trust is earned, not assumed and observability is the path to earning it.

Bio :
Mythili is a distinguished AWS Solutions Architect and cloud technology leader with over 15 years of experience across enterprise architecture, cloud innovation, and applied AI. Her technical expertise spans Java/J2EE, AWS, Azure, and Business Process Management — a foundation she now applies to designing and operating serverless and AI/ML systems in production at Amazon Web Services. A Senior Member of IEEE, IOASD Royal Fellow, IETE Fellow, AWS-certified Professional Solutions Architect, published researcher, and active IEEE peer reviewer, Mythili brings both technical depth and real production experience to the work of building AI systems that are trustworthy, observable, and built to last.