AI Exploration in Large Collections, Archives, and Libraries

This work grew out of my involvement in the Internet Archive’s digitization partnership with UC Berkeley’s Institute of Governmental Studies Library. While participating in weekly operations meetings and working with large archival collections, I began independently experimenting with emerging AI tools. The goal was to better understand how they might change the way the public interacts with digitized materials.

My approach to this work is not as a specialist, but as an observer and practitioner. The focus is on building small-scale prototypes and multi-agent systems to test the functional reality of these tools, identifying where they expand access and where practical limitations appear for researchers, institutions, and the public.

These experiments have led to ongoing conversations with university leaders, librarians, engineers, independent researchers, and the public about the opportunities and challenges of using AI with large collections of information. My goal is not to build production systems, but to understand how these tools function in real-world environments.

Below are a few of the experimental systems that came out of that work:

Multi-Agent System for Publication Intelligence, Routing, and Evaluation (MASPIRE)
A multi-agent system designed to evaluate, summarize, and classify complex documents according to institutional standards at scale.

Multi-Agent System for Polymorphic Threats (MASPT)
A simulation environment for testing how AI systems respond to evolving or shape-shifting threats.

Multi-Agent System for Subsea Threats (MASST)
An experimental system designed to detect and respond to subsea anomalies such as pressure drops, signal loss, or unusual patterns.

Additional details and prototypes available upon request. For those interested in going deeper into the thinking behind this work, see When Agents Collaborate, a blog series exploring how multi-agent systems work and what early adoption may look like for organizations and decision-makers.