Mythili Annamalai Sekar

Title of the Talk: The Right Model for the Right Problem: Choosing AI That Actually Works

Abstract :
AI observability is the practice of monitoring and understanding how AI systems behave in real-world environments, ensuring that predictions are reliable, safe, and aligned with business goals. While model development often focuses on benchmarks and test data, the true challenge lies in deployment, models can drift, fail silently, or produce unexpected outputs, especially in complex or generative AI systems.
This keynote will explore why AI observability is critical for enterprises, highlighting the risks of ignoring it and the business and operational value of robust monitoring. We will discuss the key challenges, including tracking meaningful signals across pipelines, detecting drift and hallucinations, and integrating observability into existing AI workflows. The talk concludes with best practices for enterprise AI observability, because in production, trust is earned, not assumed.

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.