Development and fine-tuning of AI models and LLMs specifically tailored for data mining in the physical sciences and engineering, Fine-tuning and evaluation of open-source LLMs (e.g., Teuken 7B, Llama 3, Mistral 7B) specifically for physics and materials science applications scenarios as well as for related research data management tasks, Developing LLM-driven agents that integrate domain knowledge to support structured chain-of-thought processes, Benchmarking and validating models against domain-specific tasks such as retrieval, summarization, metadata extraction, and structured problem-solving, Investigating advanced strategies for model adaptation, including in-domain pre-training, domain adapters, task-specific instruction tuning, and retrieval-augmented generation (RAG), Collaboration with interdisciplinary teams to enhance model performance, adaptability, and usability within physics and materials science communities, Publication and dissemination of research findings in scientific journals and conferences
