Speak at The Fifth Elephant 2026 Annual Conference

The hardest failures in our multi-agent platform didn’t come from the agents - they came from the machinery we built to keep our agents honest. This is a field report from building and operating that orchestration layer in production, where one user-turn fans out across multiple independently deployed agents, gets planned, re-planned, and critiqued before the user sees a token. We’ll set the real problem and the hard constraints, then go deep on the architecture and more importantly, the failure modes that reshaped it.

The core of the talk is trade-offs and scars, and not theory. We’ll walk through orchestration choices that we made (and ones we rejected), then spend most of the time on what actually broke under real traffic, including a few failures that inverted our assumptions about where the risk of a multi-agent system really lives.

Takeaways

If you are moving from a single-agent prototype to something multi-agent, the goal is to save you a few of the mistakes we made: which complexity is actually worth taking on, and what tends to break first.

Audience

Engineers and architects planning to build a multi-agent system

Speaker Bio

Riya is a Generative AI Engineer at Nutanix. She designs and implements intelligent systems that combine LLMs with multi-agent orchestration, tool and agent calls, and enterprise knowledge sources to deliver accurate, context-aware answers.

Draft Slides

https://docs.google.com/presentation/d/1rW-u-JwFkgtljAk-htgmOJCfxShvCckOHhiZZrKrJd4/edit?usp=sharing

Elevator Pitch

To be added