Projects & systems work
A mix of the technical and the professional, from library platforms I administer to things I've built from scratch.
Enhancing MARC serials holdings with AI
An exploratory project (and the subject of an upcoming conference poster) applying AI to a stubborn cataloging problem: turning free-text serials holdings (human-written MARC 866 "textual holdings") into structured, machine-actionable 853/863 fields. I compared approaches, including asking an LLM to read the statements directly. The one I chose put AI into building the tool rather than running it: the Holdings Workbench, developed with an AI coding assistant, is a Python application that runs deterministically, with no model or API key involved. It groups a file's 866 statements into recurring patterns, has a cataloger confirm what each captured value means, and writes the matching 853/863 fields. Whatever it leaves out, it names, and a statement it cannot fully account for keeps its 866.
A proof of concept, under active development, not a supported service.
matthewcodey.com
This very site: a Flask application I wrote, version-controlled with Git, and deployed on my own Linux server behind nginx with a TLS certificate I manage myself. A small, honest demonstration of the web-development side of the MLIS.
PNX Lookup
A staff tool that pulls a record's full normalized PNX from an Ex Libris Primo catalog by MMS ID, with no API key needed, by driving a headless browser to read the record's showPnx view, then laying it out in clean tables with CSV/Excel export. Runs locally.
Alma & Primo administration
Normalization jobs, batch processing, and Analytics dashboards and reporting across a large-scale discovery environment, plus EZproxy administration and ongoing quality control of the bibliographic database.
Accessibility workshop
Designed and delivered an accessibility workshop for library staff as part of the Accessibility, Belonging & Community committee, translating standards into practical, everyday practice.