CIR & CIS: Design Patterns for Agentic AI Systems: Lessons from Distributed Systems
Abstract: Agentic AI systems represent a shift from stateless model inference to long-running, autonomous, and distributed intelligent entities that reason, coordinate, and act within complex environments. This session frames Agentic AI as a systems design problem and draws on established principles from distributed systems to introduce reusable architectural patterns for building reliable and governable agentic systems. Key patterns—including planner–executor separation, consensus-based decision making, event-driven agents, and shared memory with eventual consistency—are examined through real-world system behaviors and failure modes. The discussion highlights how these patterns improve robustness, observability, and human oversight while mitigating risks such as hallucination propagation and uncontrolled autonomy. The session concludes by outlining open research challenges in verification, consistency, and testing of emergent agent behavior.
Date and Time
Location
Hosts
Registration
-
Add Event to Calendar
- 1111 Engineering Drive
- Boulder, CO , Colorado
- United States 80309-0422
- Building: 422 UCB
- Room Number: Third Floor, ECNT 312 Conference Room (B_ECNT)
- Click here for Map
- Starts 07 March 2026 07:00 AM UTC
- Ends 12 March 2026 02:00 AM UTC
- No Admission Charge
Speakers
Subbarao Pydikondala
Design Patterns for Agentic AI Systems: Lessons from Distributed Systems
Abstract: Agentic AI systems represent a shift from stateless model inference to long-running, autonomous, and distributed intelligent entities that reason, coordinate, and act within complex environments. This session frames Agentic AI as a systems design problem and draws on established principles from distributed systems to introduce reusable architectural patterns for building reliable and governable agentic systems. Key patterns—including planner–executor separation, consensus-based decision making, event-driven agents, and shared memory with eventual consistency—are examined through real-world system behaviors and failure modes. The discussion highlights how these patterns improve robustness, observability, and human oversight while mitigating risks such as hallucination propagation and uncontrolled autonomy. The session concludes by outlining open research challenges in verification, consistency, and testing of emergent agent behavior.
Biography:
Subbarao Pydikondala
- AI Technology Strategist
- Senior Member, and IEEE Denver Computer Society Vice Chair
Mr. Pydikondala is recognized for driving innovation in the enterprise domain and has presented at industry-leading events, including Dreamforce, AI Summit New York focusing on Agentic AI, and its safe adoption in enterprise workflows. Mr. Pydikondala’s work bridges strategic vision and real-world implementation, shaping how organizations embrace AI with digital agility. In addition, Mr. Pydikondala actively contributes to the IEEE Denver Computer Society, mentoring emerging tech talent and advancing community learning.
Address:Colorado, United States