The project aims to create models for question answering on local data by exploring generative question-answering models. Generative models can create new responses based on what they have learned as opposed to simpler "answer extraction" models which extract answers from a given context. To make these models work best with our specific local data, we need to fine-tune existing models, hence the need for GPU time. We also plan to test a RAG solution to combine LLMs and local knowledge from different databases. The PI has a PhD in language modelling, and is employed as a research engineer at the Humanities Lab in Lund. The PI is involved in other NLP projects focusing on Named Entity Extraction and Relation Extraction on medical texts.