Visual analytics · AI fairness · Model interpretability · AI safety
I build visual and interactive tools for understanding models, data, and decision processes.
I'm a Computer Science Ph.D. student at Stony Brook University, advised by Klaus Mueller. My projects range from visualizing how representations change inside language models, to systems where stakeholders negotiate dataset fairness, to dashboards that show the uncertainty behind imputed public-health data. My current focus is extending Semantic Pathway toward activation monitoring and mechanistic interpretability.
Making intelligent systems visible, inspectable, and steerable.
Recurring questions
The questions behind the projects
Across different subjects — language-model internals, causal decision processes, suppressed health data, groups changing over time — I keep returning to the same questions. What is the system or dataset actually doing? How can its behavior be represented so people can read and compare it? Where do uncertainty, bias, influence, or disagreement enter? How can someone test a change and see its consequences? And how should the result — including the process that produced it — be evaluated? Not every project touches all five, but every project starts from at least one.
Before the Ph.D., I spent five years at SAP Labs in India and Germany building production computer-vision and automation systems, with 10+ US patents from that work. It's why my research output is usually a working system, not only a paper.
Selected research
Featured projects
Four projects that show the range — from model internals to multi-agent evaluation. The research page has all projects with more detail.
Now