Catarina G Belem

I am a 5th-year PhD student in Computer Science at the University of California, Irvine, advised by Padhraic Smyth and Sameer Singh. My research focuses on building and evaluating reliable language models, with an emphasis on understanding when and why models fail—and how to make them more trustworthy. I work across several facets of reliability, including evaluation methodologies, bias analysis, and uncertainty quantification and communication. My recent projects examine gender bias in large language models, hallucinations in multi-document summarization, and how models interpret and act under varying forms of linguistic uncertainty.

Beyond academia, I have worked on real-world applications of reliable NLP through research internships at Megagon Labs and Capital One, where I focused on LLM customization for more readable and factual outputs. Prior to starting my PhD, I spent two years as a research data scientist at Feedzai, working on responsible AI in the context of fraud detection, including algorithmic fairness and explainable AI. Across both my academic and industry work, my goal is to develop principled, practical approaches that improve the transparency, robustness, and trustworthiness of modern language technologies.