First DHInfra.at LLM Workshop at DHd2026

On 23 February 2026 we ran our first hands-on LLM workshop at DHd2026 at the Universität Wien.

Eight teams of two worked through three Jupyter notebooks on OCR post-correction of the Habsburg Schematismus — a flagship use case from the Department of Digital Humanities at Universität Graz. The tasks covered the limits of prompting, the risks of fine-tuning on small data, and the benefits of synthetic data augmentation in low-resource DH settings.

Feedback was very positive; we already have requests for follow-up workshops. The event was a strong demonstration of DHInfra.at providing GPU-backed LLM and AI capacity for (DH) research in Austria.

Materials

• Notebooks, data & setup: github.com/dhinfra-at/workshop-ocr-postcorrection
• Slides: Zenodo DOI 10.5281/zenodo.18787025

Thanks to the DHInfra team, IT Services at Universität Graz, Technische Universität Graz, Austrian Scientific Computing (ASC), EuroCC Austria, and everyone who made this possible.