//Abstract In this presentation, we will explore how intelligent autonomous multi-agent systems can augment workflows. By leveraging collaborative multi-agent AI systems, people can automate routine tasks and streamline complex processes. We will go over the architecture of building multi-agent systems, talk about how to coordinate teams of AI agents that work together, discuss how to monitor and optimize these systems to be intelligent, and showcase real-world applications that highlight their potential to enhance efficiency. //Bio Natan has experience working as a Data Scientist / Software Engineer within Deloitte's Applied Artificial Intelligence division. At Deloitte, Natan collaborated on many AI projects in the domains of Natural Language Processing, Computer Vision and Big Data Analytics. He wrote the Deloitte Prompt Engineering Guide, and led execution for Ready AI, enabling clients to practically go from zero to one on their AI journeys. . This is a bi-weekly "Agent Hour" event to continue the conversation about AI agents. Sponsored by Arcade Ai (https://www.arcade-ai.com/)

//Abstract In this talk, Samuel will go into more detail on why they built PydanticAI and what problem they're aiming to solve. He'll also cover some of the future enhancements they plan for PydanticAI. //Bio Python and Rust engineer. Creator of Pydantic and Pydantic Logfire. Professional pedant. This is a bi-weekly "Agent Hour" event to continue the conversation about AI agents. Sponsored by Arcade Ai (https://www.arcade-ai.com/)

Given the popularity of generative AI, Large Language Models (LLMs) often consume hundreds or thousands of GPUs to parallelize and accelerate the training process. Communication overhead becomes more pronounced when training LLMs at scale. To eliminate communication overhead in distributed LLM training, we propose Domino, which provides a generic scheme to hide communication behind computation. By breaking the data dependency of a single batch training into smaller independent pieces, Domino pipelines these independent pieces of training and provides a generic strategy of fine-grained communication and computation overlapping. Extensive results show that compared with Megatron-LM, Domino achieves up to 1.3x speedup for LLM training on Nvidia DGX-H100 GPUs.