Designing Autonomous
AI Agents for Next-Generation Scientific Workflows

About the Session

Artificial intelligence is undergoing a major paradigm shift, transitioning from passive analysis models to autonomous, cooperative AI agents that can perceive environments, make decisions, and drive complex workflows. In scientific and technical computing, this shift represents a massive opportunity to accelerate discovery and "close the loop" between hypothesis generation, simulations, and physical experimental validation.

In this session, Kyle Chard, a leading expert in distributed systems and research automation from the University of Chicago and Argonne National Laboratory, will present a high-level, comprehensive introduction to AI agents for scientific computing, This talk will demystify what an agent is, showcase why they are vital for the future of research, and outline design principles for building scalable agentic systems across cloud and high-performance computing (HPC) environments.

What We Will Cover:

• The Anatomy of an Agent: The core fundamentals of what AI agents are, their stateful orchestration, and how they differ from traditional AI models 

• Accelerating Scientific Workflows: How cooperative agents can automate complex pipelines, from materials discovery to astrophysics, and manage distributed research data 

• Designing Agentic Architectures: A high-level framework for designing autonomous agents and agentic workflows that scale across heterogeneous and distributed computing infrastructures 

• Next-Gen Research Automation: Show how "agentic discovery" is closing the loop of scientific inquiry, shifting human effort from tedious manual execution to high-level objective setting 

• HPC and Cloud Integration: The challenges of integrating agentic workflows with high-performance computing systems and data repositories, emphasizing the need for robust middleware.

Whether you are a computational scientist, researcher, or systems architect, this session will provide you with the foundational knowledge required to design, deploy, and manage autonomous agentic workflows that accelerate innovation.