Preetam K Dutta, PhD
I am the co-founder and CEO of MarutAI, focused on building secure, production-grade machine learning systems. My work sits at the intersection of research, infrastructure, and real-world deployment.
Biography
Preetam Dutta is the founder of MarutAI, building a full-stack AI team delivered via a platform for businesses looking to reduce manual workflows or ship AI into their products.
Founded and formerly CEO of Elpha Secure, raising $28M+ from institutional investors including Canapi, AXIS Capital, and State Farm Ventures. Along his entrepreneurial journey, Preetam has invested in founders he believes in across stages including Discord, Mercury, xAI, OTOY, and GrubMarket.
Preetam is first and foremost an operator with a strong finance background and educational pedigree in AI from leading institutions. He holds a PhD from Columbia University in AI and cybersecurity and is a Yale graduate (Chemical Engineering; Economics & Mathematics).
He also serves as a National Board Member at Schools That Can, supporting access and opportunity through STEM education.
Speaker
Preetam speaks about how leaders make good decisions when the market is loud: what to believe, what to ignore, and what to build next. His talks translate AI from “possible” to “operational”—how to ship real outcomes, set ownership, and measure whether it’s working without hiring a massive team or chasing hype.
Audiences bring him in when they want frameworks, not slogans: how to turn ambiguity into a plan, how to treat risk (security, governance, resilience) as a day-to-day operating choice, and how to build durable advantages against well-capitalized incumbents. Expect clear tradeoffs, sharp priors, and practical playbooks—delivered with an operator’s pace and a researcher’s discipline.
Research
-
Simulated user bots: Real time testing of insider threat detection systems
Introduces “Simulated User Bots” (SUBs)—in situ automated users that emulate real behavior—to test and evaluate deployed insider-threat detection systems under realistic, repeatable conditions.
[Link] -
Privacy and Synthetic Datasets.
Examines whether synthetic datasets genuinely protect individual privacy, analyzing the legal and technical gap between statistical resemblance and true anonymization—and what that means for data sharing policy.
[Link] -
Machine Learning Based User Modeling for Enterprise Security and Privacy Risk Mitigation.
PhD thesis on machine-learning user modeling for enterprise security (including insider-threat detection) and privacy-aware risk mitigation.
[Link]
300+ citations · More research papers.