Naveen Setia: Powering AI that personalizes and protects

Naveen Setia began his professional journey on the trading desk at a top investment bank. This is where he put predictive AI into action. Working across asset classes - real estate, loans, and credit cards - in the US, Setia saw firsthand how predictive AI shaped portfolio pricing and future cash flows. Automation was another key theme. Setia had worked on over 100 automation projects, moving from human-driven orchestration all the way to agentic AI today. 

Setia’s encounters with AI, however, began long before his professional career. As a student in finance, his passion for the stock markets pushed him to experiment with logistic regression, linear models, and random forests. There were limitations to what he could do, but these first stabs at working with AI taught him a tough but valuable lesson: the gulf between theory and execution was wider than he thought. 

Over the years, his curiosity morphed into deep expertise. Across his career, including at his own startup and later at Paytm, Setia has consistently worked at the intersection of predictive and now agentic AI. 

Designing AI at Scale for Paytm

Setia leads Paytm’s AI strategy across multiple fronts, building systems that power everything from data infrastructure to fraud prevention and customer care. His role spans four key areas that together form the backbone of Paytm’s AI-led operations.

The first is the data platform, the foundation for every AI initiative. Paytm’s ecosystem runs on petabytes of real-time data, a scale that demands speed, reliability, and clarity. Setia and his team ensure data scientists, managers, and business leaders have seamless access to the data they need.

The second area is data science itself, where Setia becomes a consumer of the very datasets his platform provides. By bridging production and application, he enables smarter models and cleaner outputs that scale across Paytm’s vast user base.

“That's the best part because I am on the producer side of the data as well as the consumer side. I know all the nuances, and try to provide a clean database to my data scientists. This is where empowerment comes in: I have the data, now how can I use it to empower others?” he says. 

The third is the risk platform, a domain growing more critical as fraud techniques evolve. What once relied on static, rule-based systems has now transformed into AI-powered engines that detect and respond to threats like account takeover, card fraud, and scams within milliseconds. Designing these models for Paytm’s 100 million monthly active users means executing decisions in under 30 milliseconds, a technical challenge on a massive scale.

The fourth, and most recent, focus is customer care. Paytm is evolving from a bot-and-human mix toward AI-first experiences that personalize support, anticipate intent, and act on a user’s behalf. Voice agents and chatbots are being built not as replacements, but as extensions, delivering faster resolutions and more meaningful interactions.

“At Paytm, AI is not just a shiny word. While everyone talks about how we use AI, our founder's vision has baked AI into our operating DNA,” Setia says.  

This mindset informs every stage of the user and merchant lifecycle—acquisition, retention, monetization, support, and trust. Each stage requires AI/ML, but always with a human-in-the-loop to preserve confidence and safety. For example, AI determines whom to target with a personal loan, how to personalize platform interactions, and how to optimize support journeys. However, trust is non-negotiable: accounts must remain secure, loans must be underwritten responsibly, and customers must feel confident in every interaction.

Setia views his work as a balancing act: designing AI systems that drive growth while upholding security and trust. 

The challenges

The toughest part of scaling AI is data—predictive, agentic, or generative systems need clean, high-integrity datasets. At Paytm, Setia oversees both the producer and consumer sides, ensuring reliable data and validation by science teams. Another challenge is the rapid pace of AI evolution.

“The last three years have been crazy, just in terms of learning. This is an environment where everyone talks about AI, and every day is witness to something new. RAG, LLMs, and the agentic AI framework are evolving. How do you keep pace with that? Solving at such a scale has been a challenge for us.” Finally, attracting and retaining top AI talent remains critical.

Strengthening user and transaction security

Setia enables risk management at Paytm, focusing on protecting both users and transactions. Over the past few years, the team has developed a multi-layered approach to fraud prevention. The first layer targets unauthorized access, preventing account takeovers at login. The second layer, instrument takeover models, monitors transactions to ensure saved payment instruments like credit cards or postpaid accounts are not misused. Together, these models significantly reduce login and transaction fraud.

A recent focus has been scam prevention, addressing fraud that occurs outside the Paytm platform. By leveraging government-published data, Setia’s team identifies potential scamsters and provides explicit nudges to users, ensuring informed decisions before transferring funds. 

The evolution and impact of AI

Four key areas are experiencing rapid evolution, Setia believes. The first is hyper-personalization, tailoring experiences for millions of users and merchants across diverse use cases like payments, travel, or shopping. The second is conversational AI, where chatbots and voice assistants enhance customer interactions, reduce costs, and scale solutions. The third is AI-powered risk management, crucial for detecting and preventing fraud in an increasingly automated ecosystem. Finally, super apps integrate multiple services, leveraging AI to create seamless, end-to-end user experiences.

Setia also emphasizes AI’s broader societal impact. In healthcare, it can identify patterns for early diagnosis and new treatments, but humans must remain in the loop to mitigate risks. He also speaks about AI in education, revealing that he created an agent that generates personalized learning paths for his daughter, tailoring content to her unique style and encouraging practical, inquiry-based learning. 

AWS: A co-creator of value

AWS has been a long-standing partner in Paytm’s AI-first journey. Setia highlights its deep cloud expertise, continuous innovation, and broad AI/ML portfolio as key enablers. AWS provides scalable GPU infrastructure for model fine-tuning, while SageMaker supports building and deploying pipelines at scale.

Setia values AWS’s focus on emerging technologies and client-centric optimizations, which make advanced tools accessible and affordable. This partnership allows co-creation of value, combining agility, scalability, and innovation, ensuring Paytm can leverage cutting-edge AI capabilities while maintaining efficiency and speed across its data and AI workflows.

AI and beyond

Setia emphasizes the importance of being a lifelong learner, noting that stagnation can lead to lost motivation. He keeps pace with the rapidly evolving AI and ML landscape by reading research papers, attending conferences, and participating in specialized communities. He also leverages AI to personalize his learning, creating agents that summarize AI news and research reports in his preferred style, helping him quickly identify areas to explore more deeply.

Outside of work, Setia values time with his young daughter and family. He is also passionate about tracking global financial markets, following trends in stocks, commodities, and gold. A fan of non-fiction, he draws insights from real-world examples, drawing insights that apply to personal and professional spheres of his life.