We are seeking a Principal Applied Scientist to lead the next generation of click-through-rate (CTR) for Microsoft Advertising. This is a high-impact role responsible for advancing large-scale ranking models that power Microsoft Advertising, generating billions of impressions and revenue-critical decisions daily. You will combine deep machine learning expertise, solid engineering execution, and business intuition to modernize our prediction stack, drive model innovation, and mentor a growing team of scientists and engineers.

ML / Modeling LeadershipLead the end-to-end development of large-scale CTR and other user response signal models for Search and Display ads.Design, prototype, and ship cutting-edge ML architectures (deep models, multi-task, transformer-based, LLM-assisted, multimodal).Define long-term modeling strategy and roadmap with clear business impact.Technical & Engineering ExecutionModernize our current modeling pipelines, addressing critical technical debt in data flows, training pipelines, and inference systems.Partner closely with engineering teams to improve reliability, monitoring, and performance of distributed training and online serving.Introduce best practices for experiment design, ablations, feature validation, and productionization.Business & Product ImpactWork with PMs, monetization teams, and auction experts to translate business needs into modeling goals.Own model performance holistically: quality, stability, latency, and revenue impact.Develop frameworks to better understand advertiser value, user behavior, and marketplace dynamics.Leadership & MentorshipMentor and up-level applied scientists and ML engineers across the organization.Drive a culture of curiosity, deep system understanding, and high-quality scientific reasoning.Improve collaboration norms, documentation quality, and cross-team alignment.Innovation & ToolingLeverage and influence LLM-based tooling (e.g., agents, copilots) to improve team productivity and model development velocity.Identify opportunities to incorporate new modeling signals, architectures, or evaluation metrics.