1. How it works
Storing and analyzing key insight and operational knowledge from the experts who did the work.
Interviewees sit for 4-5 hour interviews that seek to capture as much as possible of the interviewee’s operational knowledge, key insight, and thoughts on how to rebuild or improve their federal job. The Callimachus Project records that interview, transcribes it, and, using AI, summarizes it into a text-based summary outline. That outline then goes into a database along with documents saved from various agencies to provide contextually-grounded answers to queries users input into Callimachus, our AI chatbot.
2. The Interview
How we collect key knowledge
The interview is the foundation of The Callimachus Project’s work. This is how we collect the key knowledge, wisdom, and insights from the experts who did the work. Whether it is USAID experts who designed and supervised programs that delivered life-saving aid around the world, NOAA experts who mapped and prepared for the next natural disaster, or EPA experts who worked with state regulators to ensure that emissions standards kept our air clean, each former federal employee has unique knowledge and experience to share. This knowledge will be critical in informing Hill and transition team decisionmakers in how to rebuild key government functions. Right now, Callimachus is conducting interviews conducted by the organization’s founder, but we have big plans to collect interviews at scale moving forward. Continue on to learn more.
3. The Product
Ethical, human-centered AI for good
At the heart of Callimachus’s model are two products that are under development. Key to both of these products is AI’s unique ability to collect, analyze, and query incredible amounts of information and distill that information into key insights. The first key product that Callimachus is developing is our scaling solution. We are building an AI voice agent product that will learn how best to interview former federal employees to learn key information from them in a format where the interviewees control the time and the pacing of the interview. You can only submit 20 minutes at a time? No problem. Or you can do the whole 4-5 hours in one sitting. With your unique session ID and login, you’ll be able to choose how to do your own interview.
This scaling solution will power the AI product at the heart of the vision of The Callimachus Project: Callimachus itself. Callimachus will be an AI-enabled LLM chatbot using retrieval-augmented generation to provide accurate information to its users drawn from the interviews collected by the Project. It will query our internal corpus of information (interviews and any agency-specific documents in our database) and cross-reference it with its outside knowledge set to provide maximally context-dependent, accurate answers to users’ questions. And, all the while, Callimachus is committed to building our AI products with the 6-Pack of Care in mind, an ethical AI framework that centers human oversight, feedback, and benefit.
4. The Dream
Why we do it
We hope that when the departments and agencies that have been hollowed out are rebuilt, new hires (and those rehired into their old roles) will be able to lean on Callimachus’s collection of expert knowledge to hit the ground running and provide valuable services to the American people as quickly as possible. We support uses of AI that support humans doing uniquely human work. So much of government work is about trust, relationships, and uniquely human expertise. We want our tools to support people being hired into roles where their expertise can flourish, supported by Callimachus.
But more than that, and the reason we are committed to remaining a non-profit: we want anyone who wants to learn about what federal government employees do every day to be able to do so. The more that the public learns about the key functions provided by government employees, and how those roles helped them and others around the world, the more people will support rebuilding, and retaining, those key functions.
