• Design, develop, and maintain features and feature pipelines for priority banking use cases such as fraud detection, credit risk assessment, collections, customer analytics, and operational efficiency.

• Translate business requirements into feature specifications with business teams and data scientists, define success criteria, and ensure proposed features align with approved business objectives and governance requirements.

• Perform data exploration, cleaning, transformation, and statistical analysis; handle missing values and outliers; and create meaningful features such as transaction patterns and customer activity measures.

• Apply feature selection and dimensionality reduction techniques to retain useful information, remove redundant inputs, and improve model efficiency and predictive performance in collaboration with data scientists.

• Build and maintain reliable, scalable feature pipelines with Data Engineering and MLOps teams, ensuring consistent transformation rules and feature values between model training and operational use.

• Test feature quality, accuracy, completeness, and availability; prevent data leakage by excluding information that would not be available at prediction time; and retain reproducible evidence of testing.

• Prepare complete feature documentation, including business rationale, definitions, data sources, transformation rules, assumptions, limitations, and version history, to support reuse, model review, approval, and audit.

• Ensure features and pipelines are developed and maintained in line with the Bank’s AI governance framework, model risk management standards, data governance requirements, responsible AI principles, and applicable regulatory obligations.

• Assess and address feature-related risks involving data quality, bias, privacy, and stability, and escalate material issues through the appropriate governance channels in collaboration with data scientists and control owners.

• Collaborate with Data Engineering, MLOps, IT, and data scientists to support controlled deployment, integration, monitoring, and change management for feature pipelines.

• Monitor feature pipeline stability, accuracy, and changes in input data; investigate failures and inconsistencies; and support timely remediation to maintain reliable model inputs.

• Maintain a catalogue of reusable feature sets, version control, and traceability for source data and transformation code, and report the number of feature sets developed and available for reuse.

• Measure and report improvements in model performance attributable to feature engineering, together with feature pipeline stability and accuracy, using agreed evaluation methods with data scientists.

• Apply appropriate controls for data confidentiality, customer privacy, access management, and secure handling of sensitive information, and share feature engineering practices that support responsible and evidence-based use of AI.

CRDB Commitment