Background

I grew up in a farming family in India. I studied computer science at VIT University, spent five years at SAP Labs in India and Germany, and am now a Ph.D. candidate in Computer Science at Stony Brook University, working with Klaus Mueller.

At SAP I built production systems for automating other software through the screen — computer vision and OCR for black-box UI automation, annotation pipelines, and an NLP-based support system. That work led to 10+ US patents, and to a habit that stuck: when a system misbehaves, build the tool that shows you why. Diagnosing failures in automation that interacted with software the way a person would meant constantly reconstructing what the system had seen and decided.

My Ph.D. work carries that habit into research. The systems I study now — language models, causal decision processes, suppressed public-health data, groups changing over time — are harder to reconstruct, and the people who need to understand them are rarely the people who built them. So I build interfaces into model internals, causal structures that stakeholders can edit, dashboards that show the uncertainty behind imputed estimates, and evaluation environments that record how a decision was reached. Coming from a background where paths like this one aren't the default, I care that these tools work for the analyst, the domain expert, and the student — not only the model's authors.