Applied Scientist II

Microsoft Microsoft · Big Tech · Redmond, WA +1 · Applied Sciences

The Copilot Agent & Search Quality team is at the forefront of core Copilot experiences across Microsoft products, including Copilot Chat and Search. This Applied Scientist II role focuses on adapting state-of-the-art vector search algorithms to the enterprise domain and applying advanced ML to improve retrieval quality. The role involves optimizing query understanding and retrieval algorithms, and partnering with platform teams to accelerate evaluation at scale.

What you'd actually do

  1. Research and evaluate tools, technologies, and methods from the community that can improve the quality, performance, or efficiency of M365 Copilot Chat and Search, and apply them to deliver business impact.
  2. Build deep knowledge of the M365 Copilot Search service while staying current with industry trends and advances in applied ML (machine learning). Consult with engineers and product teams to apply advanced concepts to search quality improvements.
  3. Use data to identify opportunities to apply state-of-the-art algorithms that improve M365 Copilot search quality. Apply statistical analysis to evaluate the behavior of deployed models, validate assumptions about evaluation results, and communicate insights to the team.
  4. Develop expertise in search relevance, NLP (natural language processing), and data-driven analysis, along with the relevant research literature and techniques. Use this understanding to identify, adapt, or create research-backed solutions—novel, data-driven, scalable, and extensible—that improve M365 Copilot Chat and Search quality.
  5. Build collaborative relationships with product and business groups across Microsoft, contributing expertise and technology that create business impact. This may include authoring white papers, developing and maintaining internal tools and services, and publishing research.

Skills

Required

  • Bachelor's Degree in Statistics, Econometrics, Computer Science, Electrical or Computer Engineering, or related field AND 2+ years related experience
  • Master's Degree in Statistics, Econometrics, Computer Science, Electrical or Computer Engineering, or related field AND 1+ year(s) related experience
  • Doctorate in Statistics, Econometrics, Computer Science, Electrical or Computer Engineering, or related field
  • equivalent experience

Nice to have

  • Bachelor's Degree in Statistics, Econometrics, Computer Science, Electrical or Computer Engineering, or related field AND 5+ years related experience
  • Master's Degree in Statistics, Econometrics, Computer Science, Electrical or Computer Engineering, or related field AND 3+ years related experience
  • Doctorate in Statistics, Econometrics, Computer Science, Electrical or Computer Engineering, or related field AND 1+ year(s) related experience
  • equivalent experience
  • 1+ year(s) experience creating publications (e.g., patents, peer-reviewed academic papers)
  • Experience applying machine learning, NLP, or information retrieval techniques to search or recommendation problems.
  • Experience using data analysis and offline/online evaluation to diagnose model behavior and improve product quality.
  • Applied ML engineering experience using Python and modern ML frameworks such as PyTorch.
  • Hands-on experience with dense/vector retrieval and text embeddings, including bi-encoders, approximate nearest-neighbor search, contrastive training, or hard-negative training.
  • Familiarity with information retrieval fundamentals and retrieval evaluation, including metrics such as recall@k and nDCG, offline evaluation pipelines, and interpretation of A/B test results.
  • Experience fine-tuning or adapting transformer or large language models for a specific domain, including domain adaptation, distillation, parameter-efficient fine-tuning, and curation of clean in-domain training data.
  • Solid applied ML engineering skills in Python and PyTorch, with experience running large-scale training and evaluation jobs on cloud or GPU clusters.

What the JD emphasized

  • state-of-the-art vector search algorithms
  • advanced ML
  • search quality improvements
  • state-of-the-art algorithms
  • applied ML
  • search relevance
  • NLP
  • data-driven analysis
  • research-backed solutions
  • novel, data-driven, scalable, and extensible
  • machine learning
  • NLP
  • information retrieval techniques
  • data analysis and offline/online evaluation
  • dense/vector retrieval and text embeddings
  • approximate nearest-neighbor search
  • contrastive training
  • hard-negative training
  • information retrieval fundamentals
  • retrieval evaluation
  • fine-tuning
  • transformer or large language models
  • domain adaptation
  • distillation
  • parameter-efficient fine-tuning
  • curation of clean in-domain training data
  • large-scale training and evaluation jobs

Other signals

  • adapting state-of-the-art vector search algorithms to the enterprise domain
  • applying advanced ML to improve the retrieval quality of business-critical information
  • jointly optimizing query understanding and the retrieval algorithm
  • accelerate the evaluation of retrieval algorithms at scale