At NexDiscovery, I develop and benchmark machine learning models for business relationship classification. I also build NLP pipelines that extract relationships and entities from unstructured text, turning business data into structured insights.
At Control Mapping, I built a coreference-enhanced entity extraction workflow using FastCoref and spaCy NER, improving organization name linking accuracy by 40%. I also developed an automated regulation extraction pipeline that reduced manual compliance review time by 60%.
My independent research evaluated how linguistic vagueness affects transformer-based sentiment classifiers, finding prediction stability above 95% for DistilBERT and RoBERTa. I work with Python and tools including PyTorch, HuggingFace Transformers, and spaCy.

