We turn everyday, messy queries into precise, actionable signals that power M365 Copilot and Microsoft Search. LU sits on the critical path—it can run on every query—classifying intent and slots and shaping the downstream search plan across providers in the substrate stack. Our work ensures that all tool calls are appropriately scoped and parameterized so Copilot starts with the right grounding data, improving relevance and limiting hallucinations.
As Senior Applied Scientist - Language Understanding and Grounding Data Quality, what you’ll work on
LLM/SLMpowered LU: intent & slotting with calibrated confidence, optimized for low latency and high reliability in Copilot experiences
MultiQuery (MQ) expansion: embrace ambiguity by exploring multiple interpretations and smarter fallbacks to boost recall without sacrificing precision.
Finetuning & eval tooling: Our team works to improve model quality specific to tool calling: Making sure the right tool is called in the right way at the right time to get the best grounding data possible. To do it, we post-train LLMs iterating fast with robust datasets, prompt evaluation, and shiproom discipline to raise quality week over week through leading edge fine-tuning
Global reach: help bring LU to more languages and markets with compliant modeling practices and scalable i18n strategy.
Deep partner collaboration: work side-by-side with Relevance and Reasoning on unified strategies that blend lexical + semantic retrieval.
Microsoft’s mission is to empower every person and every organization on the planet to achieve more. As employees we come together with a growth mindset, innovate to empower others, and collaborate to realize our shared goals. Each day we build on our values of respect, integrity, and accountability to create a culture of inclusion where everyone can thrive at work and beyond.