UX researcher and designer working at the intersection of human-centered AI, conversational systems, and experience design.
Let's design AI systems people can understand, trust, and control.
Cleo is a conversational product advisor I designed and built to address a core problem in LLM-based search systems: opacity. A hybrid architecture separates a deterministic, auditable ranking mechanism from a constrained language model, so every recommendation can be traced back to catalog evidence instead of being generated freely. The result: transparent reasoning, controllable comparisons, and decision support without sacrificing the fluency of natural conversation.
Hybrid ranking + constrained generation makes chatbot recommendations auditable, not just fluent.
Novices benefit most from combined explanations and performance categories; experts are unaffected either way.
Natural-language explanations improved perceived transparency and ease of understanding a recommendation, but didn't move decision confidence or purchase intent. Highlighting alone had no measurable effect.
Explanations were valued in principle but caused the single most severe usability problem in a moderated think-aloud study of Cleo.
Identified nine distinct product-search strategies across 31 shoppers, mapped to information-seeking models.
AI-generated SERP summaries can lower mental workload in academic search, but effects are context-dependent.
LLMs reduced workload for dataset discovery, but users stayed cautious about verifying results.
AI-generated summaries were the most understandable enrichment strategy and increased sharing of scientific posts.
My path runs from media design through UX design to human-computer interaction research. Each step sharpened the same instinct: turn what people need into concepts, flows, and decisions a team can act on. Lately that instinct is pointed at conversational AI: designing and evaluating chatbots and LLM-based systems that stay legible and controllable, not just fluent. I currently work as a Research Associate and Doctoral Candidate at GESIS – Leibniz Institute for the Social Sciences.