He came to the problem from the quantitative side. He earned his PhD at MIT, where his research at the MIT-IBM Watson AI Lab focused on causal inference and using observational data to determine what actually drives an outcome.
He later brought that work into industry, building quantitative alpha models at Point72 and training frontier AI models at Scale AI. Across research, finance, and AI, the common thread has been the same: separating signal from noise and building models that can hold up when the answer matters.
That work is now the engine that decides what a revenue or profit decision is worth before it is made.
Read the dissertation →Methods from the thesis that run in the causal engine.
First author, American Journal of Epidemiology, 2025. Whether a diabetes drug prevents cancer, across 93,353 patient records.
Accepted at ICML 2026, Seoul, one of the three top-tier machine learning conferences.
Read the paper →Contributor to DoubleML and EconML, the reference libraries for causal machine learning.
He came to the problem from inside the rooms where major business decisions get made. His career began at Deloitte, building CRM systems and learning how valuable, and messy, a company’s own customer data can be.
He later founded a venture-backed machine learning company that used credit card processing data to build customer analytics for merchants. The company was acquired.
Over the next fifteen years, he worked directly with 400+ CMOs and marketing leaders across companies including L’Oréal, Sephora, Mattel, Amorepacific, and Balmain. He kept seeing the same thing: companies making million-dollar decisions on price, promotion and spend with plenty of data, and no way to know which of them would actually make money.
Kapnova was built to find that money and put a number on it.
CMOs, directly, across fifteen years.
CPG and consumer brand companies.
Companies founded, one venture-backed and acquired.
The realization was simple and it was the whole company. The methods trusted to weigh a cancer-drug question are the methods a revenue decision needs and almost never gets. One of them had the method. The other had the problem, and the people who live it. They built the first product together on the MIT campus, where the two had met. Kapnova finds the revenue and profit a business is not booking, and puts a number on each decision that could recover it, without anyone hiring a quant desk to run it.