Dr. Rakeshnag Dasari

Building Acalable and Secure AI Agents

Abstract:

As autonomous agents transition from research laboratories into production systems at scale, we face unprecedented architectural challenges requiring both technical rigor and visionary thinking. This keynote addresses the critical question facing AI teams: how do we build agent systems that are simultaneously massively scalable and rigorously secure?

Drawing from distributed systems research and real-world production deployments, this 11-minute presentation explores three core dimensions essential for frontier-ready agent architectures. First, scalability: moving from managing dozens of agents to thousands through hybrid stateless/stateful architectures, gossip-based coordination, and observability-driven design. Second, security: implementing capability-based permissions, adversarial robustness, and cryptographic audit trails that make agents trustworthy in regulated environments. Third, practical implementation: a three-layer framework (orchestration, memory infrastructure, observability) that senior engineers can adopt immediately in production systems.

Beyond current best practices, the keynote points toward unsolved frontier challenges that will define the next decade: self-improving agents that enhance themselves safely, predicting emergent behaviors in multi-agent systems, formal verification for trustworthy autonomous decision-making, and cooperative learning frameworks where multiple agents improve together.

Attendees leave equipped with actionable architectural patterns and inspired to tackle the hard problems shaping AI infrastructure's future.

Profile:

 

 

 

I am a Sr. DevSecOps / AI/MLOPs Engineer with over 13 years of experience specializing in MLOps, cloud automation, and secure infrastructure on Azure, AWS, and GCP. Holding a Ph.D. in Computer Science and Engineering, I focus on building scalable machine learning operations pipelines integrated with observability-driven AI to enhance system intelligence and reliability. My keynotes share insights on emerging MLOps trends, automation strategies, and leveraging AI/ML with observability to drive innovation, optimize performance, and ensure operational excellence in enterprise cloud ecosystems.