Agentic Memory Management
How do you make sure your AI agents don't forget important details, get confused by user inputs, or just agree with whatever you say instead of giving you the right answer?
In this session, we will look at how AI memory actually works. We'll explore why popular memory setups can sometimes make models less accurate, and how to build simple, smart memory structures that keep your agents on track.
Agenda:
a. How AI Remembers: A simple breakdown of context windows vs. long-term memory.
b. The 'Yes-Man' Problem: Why AI agents sometimes agree with your mistakes instead of correcting you, and how to stop it.
c. Keeping Data Clean: Simple strategies to prevent user inputs from confusing your agent's memory.
d. Smart Memory Setups: How to connect external databases so your agent has access to a source of truth.
e. Live Test: Testing how your agent remembers details across a multi-turn chat.
About the Speaker:
Dr. William J Leon is an AI+XR founder and Lecturer at Cornell Tech, where he teaches Producing Agentic Systems and AI for Urban Data. He specializes in building custom augmented workflow systems for aerospace, precision manufacturing, and biotech. He holds a Ph.D. from Lehigh University and an MBA from Cornell Tech.