The University of Zambia is working to improve the visibility and impact of its research output by automating processes for curating scholarly works. Current efforts include developing machine learning models to automatically classify electronic theses and dissertations (ETDs) by domain, type, and subject using features extracted from the manuscripts. Preliminary results show models that can accurately classify ETD collections, types, and subjects have been implemented and deployed as a web API. This aims to address challenges around research output not being properly archived and discoverable, making the impact of research difficult to measure.