Drimva

Microsoft

Principal Applied Scientist - Copilot Quality

$166K–$331K

Redmond, WA +1 · On-site · Principal · Full-time · Posted Oct 23, 2026

Location
Redmond, WA
Mountain View, CA
Experience
8+ years with a bachelor’s degree, 6+ years with a master’s degree, or 5+ years with a doctorate; equivalent experience is also accepted.
Education
Bachelor’s, master’s, or doctorate in Statistics, Econometrics, Computer Science, Electrical or Computer Engineering, or a related field; or equivalent experience.

About the job

Leads the strategy and development of Copilot quality systems, using customer signals, evaluations, experimentation, and machine learning to identify failures and verify product improvements.

What you'll do

  • Develop quality metrics, evaluation datasets, labeling approaches, and benchmarks grounded in customer tasks.
  • Diagnose Copilot failure modes and improve quality through methods such as post-training, retrieval and ranking, prompting, orchestration, and product changes.
  • Establish A/B experiments, quality gates, regression protection, and rollout criteria tied to customer outcomes and release risks.

What they ask for

  • Bachelor’s degree in a listed or related field and 8+ years of related experience; or a master’s degree and 6+ years; or a doctorate and 5+ years; or equivalent experience.
  • Related experience in areas such as statistics, predictive analytics, or research.

Nice to have

  • Advanced degree (MS or PhD) in a relevant technical discipline.
  • Experience improving large language models, AI assistants, agents, or other large-scale AI products; or significant experience building, evaluating, or improving production ML or AI systems.

Pay

USD $165,600 - $296,400 per year (base salary); USD $220,800 - $331,200 per year (base salary)

Tools

machine learning, A/B experimentation, post-training, retrieval, ranking, prompting, orchestration, large language models

AI focus

Improve Copilot quality and reliability through machine learning, AI evaluation, post-training, and product experimentation.

Model-written summary · the original posting is authoritative

First seen Oct 24, 2026 · Last verified Nov 5, 2026 (UTC)