From Multi-Agent Patterns to Reliable Orchestration: My Talk at the Agentic AI Summit 2026, UC Berkeley

2 min read

I spoke at the Agentic AI Summit 2026 at UC Berkeley, the world's largest event dedicated to agentic AI. It was an honor to be selected as a speaker and contribute to the event, learn from others, and connect with the AI community.

My short talk, From Multi-Agent Patterns to Reliable Orchestration, covered multi-agent coordination patterns, how agent delegation differs from tool selection, ambiguity in delegation decisions, and orchestration as a sequence of those decisions.

My Talk Recording

From Multi-Agent Patterns to Reliable Orchestration

My Slides

Open Slides in Browser

Click into the slides and use the arrow keys, space, or swipe to navigate.

A Block I Recommend: Agentic AI Foundational Capabilities

As an attendee, one block I particularly enjoyed was Agentic AI Foundational Capabilities.

  • Dawn Song (Professor, UC Berkeley; Co-Director, Berkeley RDI; VP of AI Research, Meta Superintelligence Labs) - Towards Building Safe and Secure Agentic AI (watch at 00:30)
  • Wojciech Zaremba (Co-Founder, OpenAI) - Building Resilience for the Intelligence Age (watch at 23:55)
  • Jerry Tworek (CEO, Core Automation; Former VP of Research at OpenAI) - Opportunities and Challenges for Long-Horizon Agents (watch at 40:15)
  • Oriol Vinyals (VP of Research, Google DeepMind) - Recursive Self Improvement... of What? (watch at 50:30)
  • Dan Roth (Chief AI Scientist, Oracle; Professor, UPenn) - AI for Data and Data for AI (watch at 1:03:30)
  • Weizhu Chen (Technical Fellow & CVP, Microsoft AI) - Continuous Model Improvement (watch at 1:16:50)

The block closed with a panel bringing all six together, moderated by Maxwell Zeff (Senior Writer, Wired) (watch at 1:27:30).

This recording starts with the block above and covers roughly its first 2 hours.