Juan-Francisco Reyes

• Researched Explainable AI in LLMs: BERT, SHAP, others. • Led research on NLP and LLMs, resulting in 1 PhD thesis, 4 MSc theses, and 3 published papers on computational linguistics. • Supervised a team of 9 students to build 4 NLP datasets and fine-tune 4 LLMs. • Led 18 students in developing information extraction pipelines, from data mining (web scraping) to knowledge graph population (triplets). • Developed 4 information extraction systems following the traditional pipeline(NER, RE, EL SPO triples).

• Founded and scaled a UX Design and Development agency from 1 to 14 clients within 2 years, achieving a 45% client retention rate. • Developed a goal-driven UX framework (UX Marketing) that improved client conversion rates by up to 60%, task completion by ~40%, and user satisfaction scores by ~35%. • Built and mentored a team of 5 UX professionals, fostering a collaborative environment for design. • Established a prototyping framework using JS, HTML, and CSS that cut validation time for clients by ~50%.

• Established the first UX team (4 designers) and embedded user-centered practices design into 30+ projects. • Delivered research-driven client presentations that secured 4 major accounts, contributing to ~30% yearly revenue. • Translated research into product with a design framework, aligning design/engineering/stakeholders through A/B testing and user research—cutting iteration cycles ~30% and raising client satisfaction ~25%. • Championed UX advocacy, ensuring UX principles adopted across the agency.

Graduate Assistant at Langsdale Library, University of Baltimore. At the Department of Integrated Digital Services.Graduate Assistant at Langsdale Library, University of Baltimore. At the Department of Integrated Digital Services. Skills: User Interface Design · XML · HTML5 · Web Development · JavaScript · Usability Testing

Thesis "NLP Models for Automatic Extraction of Political Statements at the Passage Level", in which I used AI explainability methods to study how LLMs understand linguistic structures (features) to make predictions. Along the process, I built a web crawler, web annotation tools, NLP models and datasets, and a seminar in NLP and Information Extraction that I lectured for four years at BTU.

Thesis: “Effect of emotion on Marketing Landing Page Conversion”, in which I ran experiments to measure the effect of emotion on landing page conversion, using physiological instruments to measure emotional arousal (eye tracking, Empatica E4 wristband EDA sensor) and a self-reported tool to measure emotional valence. This research establishes a multidisciplinary method in interaction design to measure user experience in the context of UI design and marketing.