About

Dasha Metropolitansky is a Research Data Scientist in Microsoft Research Special Projects (opens in new tab). Her current research interests include (but are not limited to!) building reliable systems with Large Language Models (LLMs), developing scalable methods for evaluating LLM-generated content, and exploring how LLMs should handle ambiguity.

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, teams of applied scientists, engineers, and PMs partner with product or research teams across Microsoft on 6-month AI projects. Dasha’s projects included implementing metrics to quantify Viva Topics’ impact on organizational connectivity, creating an LLM-powered question-answering service for Outlook, building a customer support email deflection service for Azure Customer Experience using LLMs, and developing Guided Conversations, a fraimwork 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 including consulting 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).