This page lists my Applied Research projects. For Theory projects, see Publications.
Utilized OCR and LLMs to extract data from legal documents and designing a universal tool capable of processing data from 1,500 diverse documents.
Investigated the impact of OCR preprocessing on data extraction by comparing various OCR tools (incl. Adobe Acrobat, Abby FineReader, and Microsoft Azure) and evaluated multiple modern LLMs (incl. GPT-4o, Claude 3.5, and DeepSeek).
Performed the multiple prompting technique: determined the presence of information about certain parameters using the first prompt, extracted the values using the second prompt, and clarified inconsistencies using the third prompt.
Developed ML models for the ENFLATE project to predict household energy consumption, comparing various methods like time series analysis, regression, and neural networks.
Applied outlier-robust and sparse regression for improved accuracy.