Principal Applied Scientist - Foundation Models, Agents & Decision Intelligence

Overview

Microsoft Advertising is developing cutting-edge AI technologies aimed at understanding advertiser behaviors, detecting anomalies, and identifying emerging threats.

Role Description

We seek a Principal Applied Scientist with a robust background in mathematics, statistics, and core machine learning to lead innovations in foundational models for behavioral insights, anomaly detection, threat modeling, decision uncertainty estimation, and tool-using agents that assist in investigation and decision-making processes.

The role involves working with extensive behavioral, multimodal, temporal, and relational datasets to create scalable AI solutions that adapt across varied products, markets, policies, and adversarial conditions.

This position requires hands-on leadership from defining scientific objectives through model design, large-scale training, evaluation, deployment, and achieving tangible product outcomes.

Responsibilities

• Steer scientific projects focused on foundation models, behavioral analysis, anomaly and threat detection, and developing agentic AI systems.
• Build scalable learning frameworks for entity, content, relationship, and behavior analysis, identifying risks including new and evolving threats.
• Design methods to quantify and propagate uncertainties at all levels including model outputs, agent actions, retrieved evidence, and human decisions.
• Leverage uncertainty and business impact metrics to decide automation, evidence gathering, system escalation, or human review.
• Convert threat and adversarial insights into actionable data strategies, model designs, agent functions, and evaluation methodologies.
• Enhance agent training and evaluation that use tools and evidence collection to handle complex investigative tasks producing well-supported outcomes.
• Address machine learning challenges like distribution shifts, limited or delayed labels, noisy or imbalanced data, partial observations, and adaptive adversarial behavior.
• Implement scientific breakthroughs into dependable, efficient production-scale solutions within Microsoft Advertising.
• Provide technical guidance, mentor team members, and shape the AI architecture for trust and safety systems.

Qualifications

• A Bachelor’s, Master’s, or PhD in Computer Science, Mathematics, Statistics, Electrical Engineering, Operations Research, or related quantitative disciplines along with pertinent industry or research experience.
• Strong grounding in probability, statistics, linear algebra, optimization, numerical techniques, experimental design, and decision theory.
• Advanced knowledge of modern machine learning including foundation or representation learning, behavioral and temporal modeling, and anomaly detection.
• Demonstrated expertise in post-training analysis and evaluation of large-scale models.
• Experience modeling uncertainty within production-level decision-making systems.
• Skills to conceptualize and model threat and abuse scenarios effectively.
• Proficiency in Python programming and frameworks such as PyTorch, JAX, TensorFlow, or equivalents.
• Proven track record of advancing scientific ideas through experimentation, deployment, and measurable impacts.
• Demonstrated leadership in scientific domains including directing research, architectural decisions, mentoring, and cross-team collaboration.

Preferred Qualifications

• Experience with tool-using agents, retrieval processes, agent post-training, reward-based modeling, or trajectory analysis.
• Background in trust and safety domains like fraud detection, abuse prevention, cybersecurity, content moderation, account integrity, or policy enforcement.
• Familiarity with temporal, multimodal, heterogeneous, or graph-based data formats.
• Consistent scholarly or practical contributions in machine learning areas including agents, anomaly detection, probabilistic modeling, adversarial ML, multimodal learning, or safety and trust in AI.

Additional Information

The role will remain open for applications for at least five days and will continue until the right candidate is found.

Microsoft is committed to equal employment opportunities and welcomes applications from all qualified individuals regardless of personal characteristics protected by law. Reasonable accommodations are available upon request during the application process.