Building Stacebot: Lessons from Our First AI Assistant MVP at CELS

America/Chicago
1404 (Building 240)

1404

Building 240

Description

As part of the AiX Innovation Exchange, we invite you to an engaging session led by Michael Cunningham, Devin Hodge Fellow at the CELS Directorate. Michael will share the story behind Stacebot, an AI assistant built to support onboarding and helpdesk operations within the Research Computing team.

What makes this story special? Michael began with no prior experience developing AI applications. Leveraging Dify—a low-code platform for building AI tools—and the Argo API—Argonne’s internal AI resource—he was able to develop a functioning MVP to solve real-world problems.

This session is perfect for IT professionals and AI-curious individuals who want to understand what it takes to go from concept to prototype using LLMs. Attendees will walk away with:

  • A practical introduction to key LLM tuning parameters: temperature, Top-k, and Top-p

  • A beginner-friendly overview of how LLMs consume knowledge using vector databases and chunking strategies

  • Insights into structuring knowledge bases for both humans and machines using the Knowledge-Centered Service (KCS) framework

  • Tips, lessons, and real-world gotchas from Michael’s hands-on experience

Whether you're exploring AI for automation, service support, or innovation, this talk will leave you better equipped to start your own journey.

Registration
Participants
The agenda of this meeting is empty