AI Researcher at Independent
Nilanjan Kar is an AI researcher based in Wisconsin, focused on making language-model systems trustworthy enough for decisions people act on. His work addresses a core problem with generative AI: models that state facts or cite sources that do not hold up. His 2026 preprint, Deterministic Citation Validation for Grounded Language-Model Systems (SSRN), describes an architecture in which every citation a user sees is carried from a source record rather than written by the model. In its evaluation, record-carried citations verified 99.9% of the time, against 6.0% for a general AI assistant using web search. He is a co-inventor on a pending US provisional patent application covering this approach. A second 2026 study, Structured, Closed-Set Inputs for LLMs: Why Schema Alone Is Not Enough,...