I build AI systems that have to work outside a demo. A good model is not enough. The data has to be usable, the decision has to fit the workflow, and the result has to stand up to scrutiny.
I earned both an M.S. and Ph.D. in Industrial Engineering at the University of Cincinnati. My doctoral research focused on machine learning for feature selection with missing data; my earlier graduate work explored path planning for mobile robots.
At JPMorgan Chase, I became VP and Modeling Manager for Basel capital across retail auto and student lending. I led model developers in the United States and India working on decision trees, regression, scorecards, validation, and documentation for a roughly $50 billion portfolio. Earlier work there included fraud-detection tools and overdraft-collection strategies.
That career has led to four patents as a named inventor and more than 30 enterprise projects spanning pricing, credit, fraud, supply chain, and operations. The common thread is turning ambiguous business behavior into a system that can be measured, challenged, and improved.
Today I’m the founder of Phenx Machine Learning Technologies. I use this site to publish the experiments behind the work, including the negative results and measurement mistakes that polished case studies tend to erase.
Selected career
- 2021 to present
CEO & Founder · Phenx
Building AI and decision systems for businesses with demanding operational and regulatory constraints.
- 2017 to present
Principal Scientist · Veritas Automata
Bringing machine learning to medical, retail, industrial, and supply-chain work.
- 2025 to present
Co-Founder · Daitaboost
- 2010 to 2017
VP, Modeling Manager · JPMorgan Chase
Led Basel capital modeling for retail auto, student lending, and consumer banking.
- 2005 to 2006
Programmer Analyst · Cognizant
Worked as a programmer analyst before returning to graduate school.
What I care about
I care about building scalable, purpose-built AI systems for business problems where mistakes cost money or slow the operation. I do not start by asking which model to use. I start with the outcome. What are we trying to change? Which variables drive it? What gets in the way? How long does it take to see the effect of a decision?
I borrow a useful habit from physics: describe the system before trying to control it. In a business, that means looking for the constraints that bind, the delays that hide cause and effect, and the feedback loops that amplify small decisions. A pricing rule can shift demand. A credit policy can move risk. An operational bottleneck can erase gains made everywhere else.
Only then do I decide what to build. The job may call for a forecasting model, an optimization layer, or a decision service. Sometimes the right answer is no AI at all. Whatever we build has to scale, make sense to the people accountable for it, and improve a business result we can measure.