Principal Data Scientist: Real-Time Fraud & Protection

Capital One; North Dakota; On-site; USD 162,000 - 185,000; Full time

Job summary

Capital One is seeking a Principal Associate, Data Scientist to join the Bank Customer Protection Debit & Claims Data Science team in McLean, VA. You will help build real-time fraud detection models that protect customers during debit transactions, collaborating with data scientists, engineers and product managers.

You’ll work with Python, SQL, Spark, AWS, and advanced ML techniques including gradient boosting, deep learning, and graph/sequence methods, moving models from design to production.

Qualifications

• Pursuing a STEM degree with expected completion by start date.
• Minimum 3 years of ML experience for predictive tasks.
• 3+ years Python and SQL experience; production-quality coding preferred.
• Experience with gradient boosting models and real-time deployment.
• Experience with model backtesting and cross-functional projects.

Responsibilities

• Collaborate with data scientists, engineers, and product managers to deliver a secure fraud-detection product.
• Leverage a broad stack: Python, Conda, AWS, H2O, Spark, SQL to extract insights.
• Build ML models from design through training, evaluation, validation, and deployment.
• Translate complex technical work into tangible business impact.

Skills

Python

SQL

Spark

Machine learning

Kubeflow

XGBoost

LightGBM

H2O

Kubernetes

Data modeling

Education

Master’s degree in STEM

PhD in STEM

Tools

XGBoost

LightGBM

H2O

Spark

Kubeflow Pipelines

Kubernetes

Polars

Snowflake

Snowpark

Job description

Capital One is seeking a Principal Associate, Data Scientist to join the Bank Customer Protection Debit & Claims Data Science team in McLean, VA. You will help build real-time fraud detection models that protect customers during debit transactions, collaborating with data scientists, engineers and product managers.

You’ll work with Python, SQL, Spark, AWS, and advanced ML techniques including gradient boosting, deep learning, and graph/sequence methods, moving models from design to production.