For Material Science and Engineering (MSE) Machine Learning can be applied for i.e, Material Discovery Design [1], Property Prediction [2], Microstructure Analysis [3] etc. Ontologies are considered vital tools for MSE [4,5,6]. In ontology design the competency questions (CQs) are one important step to extract knowledge from domain experts to be formalised within the developed ontology. CQs are essential for the ontology evaluation process, because they are addressed using the knowledge embedded within the ontology. Here the power of large language models (LLMs) can be leveraged. LLMs can help material scientists to extract answers to their CQs directly from Knowledge Graphs (KG) without necessarily mastering sophisticated query languages. Here, we use a very recently developed MSE specific KG, the NFDI-MatWerk-KG as well as the MatWerk Demonstrator [7] and MatWerk Ontology (MWO) [8] and NFDI-Core Ontology [9,10], which serve as the context for the LLM. The objective is to respond to competency questions, which have been developed for the MatWerk-KG. Typically, these questions are converted into SPARQL [11] queries to check if their created ontology and the accompanying KG filled with relevant data meets the initial requirements. Our goal is to obtain answers for the CQs from the KG using LLMs in an automated manner. This is a challenging task due to the underlying complexities of translating natural language queries into structured data requests as we show in our work. One of the main challenges is to come up with a proper evaluation process regarding the performance of LLMs against manually crafted SPARQL queries for the CQs due to the fact of having various types of CQs and the nature of LLMs.