Principal Data Scientist - Identity and Matching Models

As a Principal Data Scientist, you will lead the development of advanced identity resolution and record-matching models that help people and businesses make more accurate, confident connections. You will own the full modeling lifecycle, from research and experimentation to production deployment, performance monitoring, and continuous improvement. Working with large, complex datasets from multiple sources, you will develop reliable solutions even when ground-truth labels are limited or unavailable. You will also shape evaluation methods, establish meaningful success metrics, and ensure model quality through rigorous testing and data validation. In this remote-first environment, you will collaborate closely with product, engineering, and data science teams while mentoring colleagues and influencing technical priorities. This role is ideal for an experienced data scientist who combines strong machine learning expertise with sound judgment, responsible data practices, and a commitment to measurable product impact.

• Applied data science experience: At least 6 years of experience in applied data science, with demonstrated results from developing and delivering models into production.

• Educational background: A bachelor's degree in a quantitative discipline is expected, while a master's degree is preferred. Equivalent professional experience may also be considered.

• End-to-end machine learning delivery: Experience taking models from initial research through cloud-based production deployment, monitoring, and post-launch performance evaluation.

• Technical proficiency: Strong Python and SQL skills, along with experience using Spark or similar distributed computing technologies to process very large datasets, including datasets with 100 million or more rows.

• Cloud and ML tooling: Practical experience with AWS SageMaker, Databricks, MLflow, or comparable platforms for model development, deployment, tracking, and monitoring.

• Model evaluation expertise: Ability to assess model quality when reliable ground-truth labels are limited, using sampling, manual review, and statistically informed estimates of accuracy and confidence.

• Analytical judgment: Ability to distinguish meaningful improvements from statistical noise and explain model performance, limitations, uncertainty, and trade-offs clearly.

• Model lifecycle management: Experience monitoring data changes, identifying model or label degradation, maintaining retraining processes, and documenting workflows for reproducibility and knowledge sharing.

• Responsible data management: Understanding of how to handle personal data appropriately, including adherence to opt-out rules, data retention requirements, and data lineage practices.

• Communication and collaboration: Strong written and verbal communication skills, with the ability to work effectively with engineers, product owners, and cross-functional stakeholders.

• Preferred expertise: Experience in record linkage, entity resolution, deduplication, graph-based methods, AI-assisted data labeling, or converting raw model scores into actionable confidence measures.

• Location eligibility: Candidates must be based in one of the following eligible US states: Arizona, California, Colorado, Florida, Georgia, Kentucky, Maryland, Michigan, Missouri, New Jersey, Nevada, North Carolina, Pennsylvania, South Carolina, South Dakota, Texas, Virginia, Washington, or Wyoming.