Machine learning engineer at FactSet.

LLM evaluation, retrieval, and agent observability.

I work on how language models are measured. At FactSet I build entity-graph enrichment pipelines over news articles, SEC filings, and broker research.

One question drives most of my work: how do we know when a model has genuinely improved? I research that at the Knowledge & Discovery Lab under Prof. Sourish Dasgupta, in collaboration with Oak Ridge National Laboratory, on DISCERN — a diagnostic framework for epistemic non-triviality in LLM-generated hypotheses.

Before that I was an AI Engineer Intern at Binocs Labs, building production RAG pipelines with HyDE and query decomposition, LLM reranking, and citations, and self-hosting Arize Phoenix for observability across agentic workflows. That work turned into merged contributions in Phoenix and OpenInference.

I taught IE406 Machine Learning as an undergraduate TA under Prof. Pritam Anand, convened the AI Club at DAU, mentored 500+ students through the AI Odyssey course, and co-organized Devolution with GDG On Campus DAIICT.