Dasha Metropolitansky

About

Dasha Metropolitansky is a Senior Research Data Scientist in Microsoft Research Special Projects. Her current research focuses on developing methods for evaluating the trustworthiness of content generated by Large Language Models (LLMs). For example, she created Claimify (opens in new tab) (ACL 2025 (opens in new tab)), a method for extracting verifiable claims from LLM outputs, and VeriTrail (opens in new tab) (ICLR 2026 (opens in new tab)), the first closed-domain hallucination detection method that provides provenance and error localization for processes with multiple generative steps.

Prior to Special Projects, Dasha was an Applied Scientist in the Microsoft AI Development Acceleration Program (opens in new tab) (MAIDAP), based in the New England Research and Development Center (NERD) in Cambridge, MA. In MAIDAP, Dasha worked with research and applied science teams across Microsoft on AI projects, including developing an LLM-powered question-answering service for Outlook and creating Guided Conversations, a framework for conversational agents that has been open-sourced through Microsoft’s Semantic Kernel (opens in new tab) and Semantic Workbench (opens in new tab).

Before joining Microsoft, Dasha studied Statistics and Computer Science at Harvard University. She was involved in a variety of projects, such as advising the City of Boston on improving election turnout; conducting statistical analysis for a Harvard T.H. Chan School of Public Health study testing whether methods from behavioral economics can nudge students into making more sustainable dietary choices; working with Massachusetts General Hospital’s Health Policy Research Center to collect and analyze U.S. physicians’ locations; and building a tool that maps personally identifiable information to data breaches on the dark web, as reported by VICE (opens in new tab).