Uber's Multi-Agent Platform
OSS AI, MCP, AI Infrastructure, LLMOps, Agentic AI
In this presentation, we will delve into Uber's perspective on integrating AI into the workplace and how it's revolutionizing our operations.
We'll provide a high-level overview of our approach to developing and deploying AI solutions internally.
A significant portion of the talk will focus on "AgentFx," Uber's in-house Agent SDK.
We'll explore how AgentFx empowers the development of intelligent agents that power key products and services across Uber, discussing its architecture, core functionalities, and the benefits it brings to our ecosystem.
Additionally, we'll go over how the stack works at runtime, utilizing an Open Source protocol in partnership with Langchain.ai.
During the talk, we will also quickly cover two key products: Uber Agent Builder, our no-code product, and Uber Agent Studio, both built in-house to help power agentic use cases at Uber.

Originally from Toronto, Canada, Jamieson Leibovitch has been at Uber for 4 years. He worked in Uber's Data platforms for 3 years before moving to Uber's AI Platform, Michelangelo.
He now helps lead the agent platform as a full stack engineer, working on Uber's Agent SDK.