Aswathnarayan Muthukishnan Kirubakaran

Title of the Talk:
From Models to Systems: Making AI Reliable in Production

Abstract
Artificial intelligence has progressed from predictive models to systems capable of reasoning, tool usage, and workflow execution. In real-world environments, the challenge extends beyond model accuracy to reliability, consistency, and control.
This keynote explores how AI systems evolve into production-grade systems through disciplined engineering practices. It highlights how structured interfaces, schema validation, observability, and policy enforcement enable dependable execution. Building on recent advances in agentic systems, the talk discusses how organizations can move from experimental AI prototypes to systems that operate safely and predictably in production.
Through practical architectural patterns, including governed execution layers and controlled interaction with external systems, the session demonstrates how intelligence can be combined with system-level guarantees. The keynote emphasizes that the future of AI depends on building systems that can be trusted to act correctly in complex environments.

Bio:
Aswathnarayan Muthukrishnan Kirubakaran is a Senior Data Engineer, AI researcher, and IEEE Senior Member specializing in production AI systems, distributed intelligence, and large-scale data platforms. He has extensive experience designing and deploying enterprise-grade data and AI systems that operate reliably in real-world environments. His work spans Retrieval-Augmented Generation (RAG), agentic workflows, and governed execution systems that integrate AI with infrastructure. He actively contributes to the research community as a peer reviewer and keynote speaker at international conferences. He is a Fellow of IAP, SAS, and IOASD, a Senior Member of IEEE, and a member at AAAI. His research and practical work focus on bridging the gap between advanced AI capabilities and production deployment, with an emphasis on reliability, observability, and policy-driven system design. He is particularly interested in building AI systems that can safely execute actions, not just generate responses.