Senior Applied Scientist - AI for Safe Changes Health Gating

Microsoft•Redmond, WA

3d•$119,800 - $234,700

About The Position

Microsoft is seeking an Applied Scientist to develop machine-learning and AI solutions that improve the safety, reliability, and operational efficiency of Microsoft 365 services, including systems supporting Microsoft 365 Copilot. The role focuses on scenario health gating: using service telemetry, change signals, experiments, and operational data to detect regressions, estimate risk, and inform automated or human-reviewed production decisions. The scientist will turn ambiguous reliability problems into measurable questions, develop and evaluate algorithms, implement production-oriented solutions, and work with software engineers and product managers to deploy and improve solutions in live environments. The position combines statistical and machine-learning methods with practical engineering judgment. It involves working with noisy, incomplete, imbalanced, and shifting data; accounting for uncertainty and asymmetric risk; and communicating evidence-based recommendations to technical and cross-functional audiences. The role also includes identifying and mitigating responsible-AI risks, designing suitable oversight and safeguards, and measuring model, system, and business outcomes.

Requirements

Bachelor's degree in Statistics, Econometrics, Computer Science, Electrical or Computer Engineering, or a related field and 4+ years of related experience (such as statistics, predictive analytics, or research); or a master's degree in one of these fields and 3+ years of related experience; or a doctorate in one of these fields and 1+ year of related experience; or equivalent experience.
Ability to meet Microsoft, customer, and/or government security screening requirements, including passing the Microsoft Cloud Background Check upon hire or transfer and every two years thereafter.

Nice To Haves

Master's degree in Statistics, Econometrics, Computer Science, Electrical or Computer Engineering, or a related field and 6+ years of related experience; or a doctorate in one of these fields and 3+ years of related experience; or equivalent experience.
3+ years of experience creating publications, such as patents, libraries, or peer-reviewed academic papers.
Experience presenting at conferences or other events in the outside research or industry community as an invited speaker.
3+ years of experience conducting research as part of an academic or industry research program.
1+ year of experience developing and deploying live production systems as part of a product team.
1+ year of experience developing and deploying products or systems at multiple points in the product cycle, from ideation to shipping.
Experience formulating ambiguous real-world problems as machine-learning, statistical, optimization, or experimental problems.
Experience implementing and evaluating machine-learning or statistical algorithms.
Experience with experimental design, baselines, evaluation metrics, statistical reasoning, and error analysis.
Experience working with noisy, missing, imbalanced, biased, or distribution-shifted datasets.
Ability to communicate technical findings, limitations, and recommendations to cross-functional partners.
Experience with anomaly detection, time-series analysis, causal inference, risk modeling, forecasting, ranking, optimization, or decision-making under uncertainty.
Experience building or evaluating AI agents, language-model systems, tool-using workflows, multi-agent systems, or agentic loops.
Experience developing evaluation datasets, synthetic data, simulations, golden sets, or human-evaluation processes.
Experience with cloud telemetry, observability, CI/CD, incident management, deployment safety, or distributed systems.
Contributions demonstrated through shipped systems, patents, publications, open-source work, or substantial internal technical impact.
Familiarity with Responsible AI, privacy, security, and compliance requirements.

Responsibilities

Translate service-health and change-safety problems into scientific objectives, hypotheses, datasets, baselines, and success metrics.
Develop and evaluate machine-learning, statistical, causal, time-series, or agent-based methods for regression detection, risk estimation, failure attribution, and gating decisions.
Compare technical alternatives and explain tradeoffs involving accuracy, interpretability, latency, scale, cost, reliability, and customer risk.
Address noisy, sparse, delayed, imbalanced, and distribution-shifted production data.
Assess relevant research and adapt promising methods to Microsoft 365 operational scenarios.
Build representative evaluation datasets from historical, synthetic, simulated, and production-derived data.
Design rigorous offline and production evaluations, including baselines, ablation studies, robustness testing, error analysis, and expert or human evaluation where appropriate.
Measure precision, recall, false-positive and false-negative rates, calibration, uncertainty, detection latency, and operational cost.
Create reproducible experimentation and evaluation pipelines.
Use production telemetry, incident outcomes, and user feedback to improve models and identify failure modes.
Implement algorithms, data-processing components, and evaluation systems in clean, tested, maintainable code.
Partner with engineers to integrate scientific components into scalable services and agent workflows.
Contribute to design decisions involving performance, reliability, diagnosability, monitoring, privacy, security, compliance, and cost.
Support staged deployment, monitoring, post-launch analysis, and iterative improvement.
Build reusable tools, libraries, and evaluation methods for additional scenarios.
Independently drive the applied-science workstream for one or more defined health-gating scenarios.
Collaborate with engineers, PMs, data scientists, service owners, and partner scientists.
Communicate methods, assumptions, limitations, results, uncertainty, and recommendations clearly.
Influence technical decisions through experimentation and scientific judgment.
Participate in design, code, model, and experiment reviews.
Share reusable practices and support peer development through knowledge sharing and mentoring.
Identify and mitigate risks involving privacy, security, fairness, unsafe automation, data leakage, and inappropriate production actions.
Design appropriate human oversight, fallback behavior, and safeguards for consequential decisions.
Track measures such as detection quality, gating effectiveness, robustness, automation coverage, operational-toil reduction, and production adoption.