I'm Abhinav Kumar, a credit strategy and data science leader and VKTR contributing author. I apply machine learning and causal inference to lending and customer decisions at a leading global bank.
My work sits in the gap where most enterprise AI stalls: between a model's output and the business decision it's meant to drive.
As Vice President of credit card acquisition risk strategy, I shape strategy across a $12 billion branded cards portfolio. I put machine learning models into production decisions, design champion/challenger experiments, and turn model scores into approval strategies with measurable P&L impact.
That view comes from two decades across consumer banking: frontline sales and relationship management, wealth and foreign exchange, card alliances, analytics consulting, and now credit risk strategy in the US.
Acquisition and approval strategy, loss and profitability trade-offs, controls and governance.
Propensity score matching and experiments that show what actually changes customer outcomes.
Churn, default and forecasting models used as inputs to real business decisions.
Leading global bank, Delaware, USA
Strategy across a $12B branded cards portfolio. Operationalizes ML models, runs champion/challenger tests, and owns controls for new strategies through launch and validation.
Innodatatics / ExcelR
Led machine learning and customer analytics engagements for global clients, including churn, default and revenue management.
Global bank, India
Progressed from relationship manager to corporate sales, head of cards acceptance and alliances, and senior manager for expat banking, forex and wealth. Built card programs with partners including Mercedes, Volvo and Shopper's Stop.
IBM Daksh (now Concentrix)
Customer service quality and complaint resolution for US financial services processes.
Mahajan, P., Arya, K., Agarwal, P. K., & Kumar, A. (2026). Journal of Intelligent Decision Making and Information Science, 3(8s), 1468–1487.
An interpretable ensemble that forecasts cybersecurity staffing shortages in financial market infrastructure, turning risk signals and HR data into hiring and reskilling decisions.
Kumar, A. (2026). American Journal of Technology, 5(6), 41–64.
Applies five propensity score matching methods to 10,000 bank customers to show how causal methods can complement predictive models in retention strategy.
A three-part series for VKTR, the enterprise AI publication, on turning unstructured narrative text into computational story graphs, from the case for the approach through the pipeline to production.
I welcome conversations about credit strategy, causal inference in finance, enterprise AI writing, peer review, speaking and teaching.