Join Microsoft Advertising as a Principal Applied Scientist to develop cutting-edge AI systems that understand advertiser behavior, identify anomalies, and mitigate emerging threats. This role focuses on advancing foundation models for comprehensive understanding, developing sophisticated anomaly and threat modeling capabilities, and creating intelligent agentic systems. You'll tackle complex challenges involving large-scale behavioral, multimodal, and temporal data to build resilient systems that generalize across diverse products and markets.

This is a hands-on, end-to-end scientific position. You will be responsible for the entire lifecycle of model development, from initial problem formulation and research to large-scale training, deployment, and measuring tangible product impact.

Define and lead critical scientific initiatives in areas such as foundation models, behavioral analytics, and agent-based systems. Develop scalable learning frameworks to model entities, content, and evolving risks. Create methods for modeling and propagating uncertainty throughout decision pipelines, informing automation and escalation strategies. Translate threat intelligence into actionable data strategies, learning objectives, and model architectures.

Advance the development and evaluation of agents capable of using tools and evidence to resolve complex cases. Address challenging learning scenarios, including distribution shifts and sparse data. Drive the translation of scientific breakthroughs into robust, efficient, and measurable production capabilities within Microsoft Advertising. Provide technical leadership, mentor junior scientists, and shape the future architecture of AI-driven trust and safety systems.

A Bachelor’s, Master’s, or Doctorate degree in a quantitative field such as Computer Science, Mathematics, or Statistics, coupled with relevant industry or research experience, is required. A strong foundation in probability, statistics, linear algebra, optimization, and experimental design is essential.

Demonstrate deep expertise in modern machine learning, including foundation models, anomaly detection, and temporal modeling. Proven experience in post-training and evaluating large-scale models and modeling uncertainty in production systems is crucial. Proficiency in Python and experience with ML frameworks like PyTorch, JAX, or TensorFlow are necessary. Candidates must exhibit strong technical leadership qualities and the ability to drive scientific ideas from conception to production impact.