Privacy Research Engineer, AI

Microsoft AI•Redmond, WA

2d•$142,800 - $304,200

About The Position

Microsoft AI is seeking a technically focused privacy professional to work directly with engineers, researchers, and platform teams developing and evaluating AI models. The role helps define how privacy is engineered into AI development, making sound data-use decisions before risks are built into systems. The position develops and applies methods to assess and measure AI privacy risks, including memorization, regurgitation, membership inference, model inversion, reconstruction, and training data leakage. It evaluates whether mitigations such as scrubbing, loss masking, aggregation, de-identification, differential privacy, synthetic data, filtering, and access controls are sufficient for particular datasets, models, training methods, and use cases. It also contributes to the design of data governance and infrastructure for compliant AI development at scale. This is an embedded engineering role rather than a traditional privacy review position. The work involves quantifying dataset and model risks, evaluating mitigation evidence, translating privacy and regulatory requirements into testable engineering requirements, and balancing model utility with measurable reductions in privacy risk.

Requirements

Bachelor's degree in Statistics, Econometrics, Computer Science, Electrical or Computer Engineering, or a related field and 6+ years of related experience; or a master's degree in one of these fields and 4+ years of related experience; or a doctorate in one of these fields and 3+ years of related experience; or equivalent experience.
4+ years of experience with privacy-preserving machine learning or data protection techniques, such as differential privacy, de-identification, aggregation, synthetic data, data minimization, sensitive data filtering, or loss masking.
3+ years of technical experience assessing privacy risks in AI or machine learning systems, including one or more of memorization, regurgitation, membership inference, reconstruction, model inversion, or training data leakage.
Familiarity with privacy requirements and emerging AI governance expectations under GDPR, UK GDPR, EU AI Act, DSA, CCPA/CPRA, or similar regulatory frameworks.

Nice To Haves

Experience developing privacy risk measurement, model evaluation, privacy red teaming, or control validation methods.
Experience translating privacy research, measurement methods, or experimental techniques into scalable engineering systems and practices.
Experience working with engineering teams to balance privacy risk, technical feasibility, product value, and data utility.
Experience building or influencing data governance systems, including lineage, provenance, access governance, retention, deletion, consent enforcement, opt-out enforcement, or dataset eligibility controls.
Experience working in ambiguous technical areas without an established standard playbook.
Experience designing or implementing privacy-enhancing technologies or privacy measurement techniques for machine learning systems.
Experience working across research and production engineering environments and translating experimental methods into scalable engineering practices.

Responsibilities

Partner with engineers, researchers, and platform teams on model training, evaluation, data preparation, and AI system architecture.
Define technical privacy requirements for model training and AI development, including requirements for datasets, storage, access, retention, deletion, opt-out enforcement, data subject rights readiness, and downstream use.
Develop and apply methods to assess and quantify AI privacy risks, including memorization, regurgitation, membership inference, reconstruction, model inversion, and training data leakage.
Evaluate how privacy risk changes across datasets, model architectures, training approaches, and privacy-preserving techniques.
Assess the effectiveness and limitations of mitigations such as scrubbing, loss masking, aggregation, de-identification, differential privacy, filtering, and synthetic data, and determine when additional controls are needed.
Develop measurable privacy criteria for AI systems, including risk thresholds, test methods, validation requirements, and evidence expectations.
Help architect privacy and data governance infrastructure for AI development, including data lineage, provenance, approved-use tracking, purpose limitation, access controls, auditability, retention, and deletion.
Translate privacy, legal, regulatory, and policy requirements into implementable and testable engineering requirements.
Build repeatable patterns that enable teams to develop models using governed, measurable, and model-ready data.
Represent privacy in technical design discussions balancing model quality, data utility, privacy risk, and engineering feasibility.