BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/Denver BEGIN:DAYLIGHT DTSTART:20260308T030000 TZOFFSETFROM:-0700 TZOFFSETTO:-0600 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:MDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20261101T010000 TZOFFSETFROM:-0600 TZOFFSETTO:-0700 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:MST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20260403T135839Z UID:01ABBC34-0225-47B7-87A6-D1E4FC82F18D DTSTART;TZID=America/Denver:20260311T180000 DTEND;TZID=America/Denver:20260311T193000 DESCRIPTION:Abstract: Agentic AI systems represent a shift from stateless m odel inference to long-running\, autonomous\, and distributed intelligent entities that reason\, coordinate\, and act within complex environments. T his session frames Agentic AI as a systems design problem and draws on est ablished principles from distributed systems to introduce reusable archite ctural patterns for building reliable and governable agentic systems. Key patterns—including planner–executor separation\, consensus-based decis ion making\, event-driven agents\, and shared memory with eventual consist ency—are examined through real-world system behaviors and failure modes. The discussion highlights how these patterns improve robustness\, observa bility\, 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 em ergent agent behavior.\n\nSpeaker(s): Subbarao Pydikondala\, \n\nRoom: Thi rd Floor\, ECNT 312 Conference Room (B_ECNT)\, Bldg: 422 UCB\, 1111 Engine ering Drive\, Boulder\, CO \, Colorado\, United States\, 80309-0422 LOCATION:Room: Third Floor\, ECNT 312 Conference Room (B_ECNT)\, Bldg: 422 UCB\, 1111 Engineering Drive\, Boulder\, CO \, Colorado\, United States\, 80309-0422 ORGANIZER:gowansj@ieee.org SEQUENCE:17 SUMMARY:CIR & CIS: Design Patterns for Agentic AI Systems: Lessons from Dis tributed Systems URL;VALUE=URI:https://events.vtools.ieee.org/m/545478 X-ALT-DESC:Description: <br /><p><strong>Abstract:</strong> Agentic AI syst ems represent a shift from stateless model inference to long-running\, aut onomous\, and distributed intelligent entities that reason\, coordinate\, and act within complex environments. This session frames Agentic AI as a s ystems design problem and draws on established principles from distributed systems to introduce reusable architectural patterns for building reliabl e and governable agentic systems. Key patterns&mdash\;including planner&nd ash\;executor separation\, consensus-based decision making\, event-driven agents\, and shared memory with eventual consistency&mdash\;are examined t hrough real-world system behaviors and failure modes. The discussion highl ights how these patterns improve robustness\, observability\, and human ov ersight while mitigating risks such as hallucination propagation and uncon trolled autonomy. The session concludes by outlining open research challen ges in verification\, consistency\, and testing of emergent agent behavior .</p> END:VEVENT END:VCALENDAR