University of San Francisco Department of Computer Science
MAGICS Research Lab Machine Learning·Artificial Intelligence

Meetings

Meetings are open to the public and we encourage anyone to join: beginners, practitioners, researchers — all are welcome.

Current schedule

TermFall 2026
DayMost weeks on Thursday
Time7:00 pm – 8:00 pm
RoomTBD
OnlineJoin on Zoom

Let us know if you would like to meet in person and we will find a room.

Next meeting

Project check in

Featured presentation: Why AI agents can complete single steps but struggle to finish a whole task — Thursday, November 19, 7:00 pm

Presenter: Aria Han

What happens at a meeting

Our meetings generally revolve around:

  • Project updates from lab members
  • Open discussion and problem-solving
  • Short tutorials on tools and techniques
  • Coverage of the latest research in our fields of study

Reading Groups

Throughout the semester, we host reading groups to explore current developments in the industry.

Why should you attend?

  • Discover topics you are interested in and want to work on.
  • Form ideas for research projects that you’re passionate about.
  • Practice your skills in presenting and speaking about research.

This week’s presentation

Why AI agents can complete single steps but struggle to finish a whole task. Presented by Aria Han.

An agent can handle one step. Carrying a whole task from the first request to a verified result is harder: the system has to keep context, coordinate its actions, recover from failures, and check that the result is what was asked for.

This talk looks at how agent systems are built to do that — through orchestration, persistent state, deterministic checks and structured feedback. It draws on research in agent reasoning and long-horizon task performance, and considers the challenges of coordinating complex work across multiple stages.

Observations from an implemented agent runner show how far this gets today, and what still breaks.

About the presenter

Aria Han

Aria Han is an AI systems architect and CTO specializing in autonomous agents, orchestration, and AI infrastructure. She studied Computer Science at the University of San Francisco and has founded three companies. Her work spans agent orchestration, persistent memory, evaluation, and the development of systems capable of executing complex tasks with minimal human intervention. She focuses on bridging the gap between what AI agents can do individually and what reliable, autonomous systems can accomplish over time.

Questions?

Contact the lab manager and they’ll point you in the right direction.