Principal Applied Scientist

The Online Forensics team sits within the Microsoft AI Experiences organization, where we protect the monetization network that funds the products people use every day. Because attackers adapt quickly, effective defense requires discovering how systems can be abused, not just investigating known incidents. We combine hands-on adversarial research, controlled experimentation, and large-scale data analysis to establish authoritative ground truth and expose weaknesses before they are exploited at scale. 

As a Principal Applied Scientist, you will discover how web-based monetization systems can be manipulated and build the experimental and investigation systems needed to study that abuse at scale. Your research will cover both exploitable weaknesses and abuse of intended functionality, including automation, identity and attribution manipulation, and multi-step economic exploitation. You will personally write code, construct safe adversarial experiments, and combine experimental evidence with large-scale telemetry to establish mechanism, prevalence, and impact. 

The scientific challenge is to distinguish adversarial manipulation from legitimate behavior under incomplete and potentially manipulated telemetry. You will develop methods to test competing explanations, determine whether experimental findings generalize to real-world traffic, and evaluate candidate signals without unacceptable false positives. You will turn validated approaches into reliable, reusable systems and evaluation methods that partner defense teams can adopt. 

You will own a focused portfolio of high-impact research problems, from hypothesis through implementation, validation, and adoption. At the principal level, you will set scientific and technical direction while remaining directly involved in coding, experimentation, debugging, and delivery. You will own the research capabilities you build; partner defense teams own production enforcement and remediation. 

Microsoft's mission is to empower every person and every organization on the planet to achieve more. As employees we come together with a growth mindset, innovate to empower others, and collaborate to realize our shared goals. Each day we build on our values of respect, integrity, and accountability to create a culture of inclusion where everyone can thrive at work and beyond. 

Starting January 26, 2026, Microsoft AI (MAI) employees who live within a 50-mile commute of a designated Microsoft office in the U.S. or 25-mile commute of a non-U.S., country-specific location are expected to work from the office at least four days per week. This expectation is subject to local law and may vary by jurisdiction. 

• Discover and validate abuse mechanisms. Form original, testable hypotheses about adversarial manipulation of web and advertising systems and emerging LLM-native monetization flows. Investigate system behavior, service interactions, and economic incentives to explain how the abuse works. 

• Personally implement adversarial experiments. Build automated clients, instrumented browser workflows, simulations, and replay harnesses under safe, authorized conditions with clear scope and safeguards. Use appropriate controls and competing hypotheses to distinguish demonstrated mechanisms from unvalidated assumptions. 

• Develop rigorous evaluation methods. Assess whether experiments represent real-world conditions, establish the provenance and quality of labels, and measure generalization, coverage, false-positive risk, and robustness to adaptation. 

• Build durable research and investigation systems. Design, implement, test, deploy, and maintain the services, instrumentation, telemetry pipelines, and reusable libraries needed for the research portfolio. Turn successful approaches into reliable capabilities that other researchers and partner teams can use independently. 

• Connect experiments to real-world evidence. Apply behavioral, graph-based, temporal, and cross-surface analysis to large-scale telemetry. Identify adversary infrastructure and evasion patterns, determine prevalence, and quantify economic exposure while stating uncertainty and limitations. 

• Establish authoritative ground truth. Correlate evidence across accounts, traffic, transactions, providers, devices, and infrastructure. Reconstruct attack chains and produce reproducible datasets and technical assessments that clearly state mechanism, provenance, confidence, scale, and impact. 

• Translate findings into stronger defenses. Produce candidate signals, labeled evaluation datasets, monitored indicators, and reproducible regression scenarios. Partner with teams that own production defenses to evaluate improvements and verify that corrective actions address the demonstrated abuse. 

• Lead scientific and technical work. Set priorities for the research portfolio, guide experimental and architectural decisions, mentor others, and influence adoption across organizational boundaries while making substantial direct research and code contributions. When material incidents arise in the portfolio, lead reconstruction and turn findings into reusable evaluations and capabilities. 

• Embody our Culture and Values.