Manager, Data Scientist

Capital One’s Model Risk Office is seeking a Manager, Data Scientist to partner with cross-functional teams and help identify, quantify, and govern model risks that affect decision-making, including risks related to Generative AI.

In this onsite role in Chicago, IL
, you will build, validate, and challenge models used in production. You will support model risk governance by working across data science, software engineering, and product teams to surface how model risks may impact business outcomes within the Enterprise Services division.

• Collaborate with cross-functional teams of data scientists, software engineers, and product managers to identify and quantify risks associated with models.
• Apply a broad technology stack to uncover insights in large volumes of multi-modal data, using frameworks and tools such as Py Torch and Hugging Face
  , orchestration and retrieval components such as Lang Chain and vector databases
  , along with LLMOps and observability platforms.
• Develop machine learning models that challenge existing “champion models” currently deployed in production and contribute to the model governance framework for the next generation of models.
• Validate diverse model types across multiple business domains, and present to executives how identified model risks could affect the business.

• Currently has, or is in the process of obtaining (with expectation of completion on or before the scheduled start date), one of the following:
  • Bachelor’s degree in a quantitative field (Statistics, Economics, Operations Research, Analytics, Mathematics, Computer Science, or a related quantitative field) plus 6 years of experience performing data analytics; or
  • Master’s degree in a quantitative field (Statistics, Economics, Operations Research, Analytics, Mathematics, or related quantitative field) plus the required experience as outlined in the posting.
• At least 1 year of experience leveraging open source programming languages for large scale data analysis.
• At least 1 year of experience working with machine learning.
• At least 1 year of experience utilizing relational or vector databases.

The role leverages Py Torch ,
Hugging Face
, Lang Chain
, Vector Databases
, LLMOps
, observability platforms, and additional tooling across cloud and programming ecosystems including AWS
, Python
, Scala
, and R. It also references areas such as GenAI and model evaluation and analytics methods including confusion matrix and ROC curve
, plus techniques such as clustering
, classification
, sentiment analysis
, time series
, and deep learning
.

The Capital One Model Risk Office focuses on safeguarding the company from model failures while improving decision-making through models, including unique risks associated with Generative AI (GenAI). The team applies expertise across statistics, software engineering, and business, and emphasizes continuous investment in future capabilities, tools, and partner relationships. Their approach also includes learning from past errors to develop more robust techniques to help prevent recurrence.

• Eligible for performance-based incentive compensation
  , which may include cash bonus(es) and/or long term incentives (LTI).
• Comprehensive, competitive, and inclusive health, financial, and other benefits supporting overall well-being.

• Innovative
  : regularly research and evaluate emerging technologies and seek opportunities to apply state‑of‑the‑art methods.
• Creative
  : bring definition to complex problems and share new ideas while working through questions to find answers.
• Technical
  : comfortable with open‑source languages and hands‑on experience developing data science solutions using open‑source tools and cloud computing platforms.
• Statistically-minded
  : build, validate, and backtest models, with experience interpreting a confusion matrix and ROC curve, plus work across clustering, classification, sentiment analysis, time series, and deep learning.
• Data-focused
  : able to retrieve, combine, and analyze data from multiple sources and structures, recognizing that data understanding is often key to effective data…