Research Engineer, Privacy and Anonymization (Mid-Level)

About the Role

Build privacy and anonymization systems that make sensitive real-world data safer to use in AI training and evaluation. You will develop detection and transformation methods, then bring them into production pipelines while preserving useful data structure and signal.

What You'll Do

• Build systems to detect PII, quasi-identifiers, credentials, and other sensitive information, and apply transformations suited to each data type and use case.
• Develop and benchmark detection methods using rules, statistical models, classifiers, and LLM-based approaches.
• Create production pipelines that anonymize raw data before downstream processing, training, evaluation, or synthetic data generation.
• Measure privacy risk and retained data utility with evaluation frameworks, leakage tests, and adversarial re-identification attempts.
• Design robust systems that handle new data sources, schema drift, unusual formats, and sensitive information in unexpected fields.
• Partner with engineering, research, operations, and customers to turn privacy requirements into practical technical safeguards.

What We're Looking For

• At least 2 years of experience building production data or ML systems in Python.
• Hands-on experience detecting, removing, or transforming PII while preserving useful data characteristics.
• Experience with information extraction, named-entity recognition, classification, or related methods for detecting sensitive or rare content.
• Experience building end-to-end data processing pipelines and benchmarking approaches across recall, precision, latency, cost, and data utility.
• Understanding of redaction, masking, pseudonymization, anonymization, and synthetic data generation.
• Experience handling schema drift, unusual formats, and edge cases; familiarity with privacy technologies such as differential privacy, k-anonymity, secure aggregation, or format-preserving encryption is valuable.
• Experience with sensitive data in healthcare, finance, security, or related fields, and with low-latency or high-throughput ML systems, is beneficial.