Applied Math Colloquium: Yifan Hu (UMBC)
Title: Human-in-the-loop AI agent workflow for teaching.
Abstract: How can we use AI agents to help prepare teaching? How can we design an agent workflow that provides high quality course material, and saves us time and energy? What is the role of undergraduate math instructors in this AI age? These questions arise when I started teaching MATH 404 this semester. In this talk, we will go over my journey of using AI agent workflow for teaching preparation, and show field-tested strategies and techniques from my MATH 404 classroom. We demonstrate well established agentic AI workflow that covers prompts, LLM contexts, sub-agents, evergreen instructions, and skills. These components are implemented and improved constantly in MATH 404 course notes, quizzes, and assignments. We will also briefly discuss how human-in-the-loop AI agent workflow is aligned with the UMBC AI Strategy report(https://ai.umbc.edu/umbc-ai-strategy/)