I study how AI systems reshape expert knowledge work — and how to make AI-assisted reasoning more reliable, transparent, and human-centered.

My research sits at the intersection of large language models, human–AI interaction, and evidence-grounded decision support. I am particularly interested in settings where experts use AI systems to interpret complex information, produce knowledge artifacts, or make sense of evidence under uncertainty. In these contexts, fluent AI outputs can be useful, but they can also conceal hidden failures: unsupported claims, broken dependencies, misplaced confidence, or subtle shifts in judgment.

My work develops benchmarks, evaluation methods, and human-centered frameworks for studying these risks. A recurring theme is that expert workflows are dependency-rich: goals, constraints, evidence, claims, methods, and revisions must remain coherent across time, across documents, and across human–AI interactions.

I currently frame this agenda as human-centered reliable AI. Rather than asking only whether a model produces a plausible answer, I ask how AI systems affect expert reasoning, evidence use, verification, trust, and accountability.

Two questions guide my research:

How do AI systems change the way experts produce, evaluate, and trust knowledge?

How can we design evaluation methods and human–AI workflows that support reliable, evidence-grounded, and accountable use of AI?

Background

I am completing a PhD at the Technical University of Munich at the intersection of machine learning, economics, and large language models. Before and alongside my doctoral work, I worked on applied AI and data science projects in academic and industry settings, and taught computational and data science at university. I also draw on training in design thinking, which shaped my interest in problem framing, expert workflows, and human-centered AI systems.

I hold an MSc in Data Analytics from University College Dublin and an MA in Banking and Finance from the University of St. Gallen.