Master's students often struggle with thesis topic selection due to a lack of inspiration and uncertainty about supervisor availability for specific research areas. This study presents TIDE (Thesis Inspiration Discovery Engine), a web-based tool that employs Large Language Models to extract structured information from institutional thesis corpora, providing students with inspiration grounded in locally available supervision. TIDE utilises OpenAI's GPT-4o-mini to automatically extract key information from past theses, including research questions, methodologies, supervisors, variables of interest, and suggestions for future research. The tool processes thesis PDFs through a systematic pipeline: text extraction, structured data extraction, and compilation into a searchable database format. To evaluate TIDE's effectiveness, the system was tested on a sample of 100 theses from the University of Amsterdam's Faculty of Social Sciences. Manual validation of 1,313 extracted data points revealed an overall accuracy rate of 89.95%. The errors are classified into three main categories: missing information (7.01%), hallucinated content (2.59%), and other errors (0.53%).This research demonstrates the potential of LLMs for structured information extraction from academic texts and provides a practical solution for improving the thesis topic selection process in higher education institutions.