Senior Applied Scientist

AI assistants are only as good as the information behind their answers. Our team builds the search and grounding technology that finds the right information for each request, across email, documents, meetings, chats and people, and connects AI-generated answers to the sources that support them. This work sits on the critical path: getting it right improves relevance, limits hallucinations and earns user trust.

We work across the full retrieval stack: understanding what a person is really asking, turning everyday, messy requests into precise queries, blending lexical and semantic retrieval, and deciding when enough evidence has been found. Modern language models run throughout, and we fine-tune them with supervised and reinforcement learning so they find better evidence with less wasted effort. Improvements are measured offline on relevance, accuracy, citation quality and speed, then confirmed in online experiments before they ship.

We are looking for a Senior Applied Scientist to help drive this work. You will take on hard, open-ended problems in search and grounding and own them end to end, from framing the problem and building the data through modeling, evaluation and deployment, while helping set the scientific direction for the team. Much of the work happens in a privacy-first environment where models are trained and evaluated without direct access to customer data, so creativity with synthetic data and evaluation design is essential.

The impact is broad: our work improves search in Outlook, Teams, SharePoint and OneDrive and grounds the answers in Microsoft 365 Copilot, reaching tens of millions of people at work. The role spans science and engineering, so you can learn everything from writing production code to training the latest language models, and everything in between. People who thrive here love turning ambiguity into robust systems, balance machine learning and information retrieval fundamentals with production pragmatism, and care about user trust, performance and clarity of outcomes.

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.

• Advance how AI systems search for information: understanding what a person needs, turning requests into precise queries, choosing where to look, and recognizing when enough evidence has been found.

• Improve the quality, coverage and reliability of grounding data, so answers are relevant, accurate and supported by the sources they cite.

• Train and fine-tune language models for search and grounding tasks, using supervised fine-tuning and reinforcement learning to improve both quality and speed.

• Design evaluation methods and reward signals that reflect what users value, such as relevance, accuracy, trustworthy citations and efficiency.

• Build training and evaluation data in privacy-preserving ways, including synthetic data generation.

• Define offline and online metrics for search and answer quality, and make sure offline gains predict what people experience in production.

• Dig into real-world failures, find root causes and turn them into model, data and system fixes that ship.

• Partner with engineering, product, ranking and research teams to take new techniques from experiment to production, across many languages and markets.

• Mentor scientists and engineers, raise the bar on experimental rigor, and communicate results and trade-offs clearly to technical and non-technical audiences.