Public data is only useful if people can understand it. While complete data descriptions have been a minimum requirement for 5 years, 44% of public datasets published on Washington’s open data portal have no description, and many of the existing descriptions could be more clear. Missing and unclear descriptions make it difficult for residents, researchers, journalists, and businesses to fully understand and use government data.
To help agencies address that challenge, a team of University of Washington graduate students, self-dubbed the “Four Musketeers,” developed an opensource AI tool that helps data publishers better explain what their data means and how people can use it. The project was completed as part of the University of Washington’s Master of Science in Information Management (MSIM) Capstone program in partnership with Washington state’s Open Data Program.
The team consisted of four University of Washington MSIM students: Wynter Lin, Danny Yue, Felix Zhao, and Julia Zhu. Together, they combined expertise in data science, information management, and emerging AI technologies to develop the solution over the course of the seven-month capstone project.
The tool uses generative AI to analyze datasets and create descriptions, column definitions and other documentation that meet Washington’s open data standards. It can also identify missing information and suggest improvements to existing content. Human reviewers are essential to the workflow. Publishers will review, edit, and approve recommendations before publishing to the portal.
The project first began taking shape during the annual capstone selection process, when student teams met with sponsors to learn about potential project challenges. WaTech’s Open Data Program Manager and project sponsor Cathi Greenwood recalled that one student stopped by the table, asked a few questions about the project, and returned about 20 minutes later with a working prototype. That early demonstration helped spark the vision for the full tool that followed. “The student team was amazing,” Greenwood said. “Over the past seven months, they built a tool that not only improves metadata quality but also helps evaluate the performance and cost of different AI models.”
Over the course of the project, the team refined the tool based on feedback from program sponsors, evaluated the cost andquality of different AI models, and collaborated with the Enterprise Data Program to prepare the system for real-world use. The state’s Open Data Librarian Kathleen Sullivan worked closely with the students to ensure the tool aligned with the needs of Washington’s government data publishers and public open data community.
The students designed and built the solution as an open-source project. The Open Data Program will test the tool with data.wa.gov publishers, with options to expand use as the tool proves its reliability.
As Washington continues moving toward a more data-informed, transparent, and accessible government, projects like this demonstrate the value of collaboration between state agencies and academic partners. By advancing tools that improve data quality, WaTech supports its mission to empower agencies, strengthen public trust, and ensure that government data is clear, consistent, and meaningful for everyone. The release of this open-source solution marks another step forward in making statewide public data more discoverable, usable, and equitable.