With a deep background in artificial intelligence and machine learning, Dominic Jainy has spent his career at the forefront of technological innovation. His work, spanning markets from the U.S. to the APAC region, focuses on a challenge many in fintech consider unsolvable: how to build ironclad fraud defenses without alienating legitimate customers. In our conversation, Dominic unpacks the layered AI strategies that have dramatically cut financial losses, the critical importance of breaking down data silos within an organization, and his vision for the escalating arms race between fraudsters and the systems designed to stop them.
Your team cut chargeback losses by 35% in three months. Can you walk me through the key components of the AI system that achieved this, and how you balanced tight security with a smooth customer experience? Please provide a specific example.
That success came from rejecting a false choice that plagues the industry: that you either have tight security that frustrates people or a smooth experience that lets criminals slip through. We refused to compromise. The core of our system is a layered defense, not a single gate. The biggest immediate impact came from implementing AI-powered liveness verification at onboarding. This alone was the catalyst for that 35% drop in chargeback losses. The key wasn’t just stopping fraud, but doing it without a spike in false positives. We achieved a 20% drop in our overall risk rate, which really strengthened our portfolio quality, all while ensuring honest customers could still sign up without any hassle.
You combine biometric verification at sign-up with continuous behavioral analysis. How does your system distinguish between a legitimate customer’s unusual behavior and a genuine fraud attempt after an account has already been opened?
That’s the crucial next layer of defense. Identity verification at the front door is essential, but it doesn’t help when a fraudster steals an existing customer’s phone. This is where our continuous monitoring becomes so critical. We analyze behavioral patterns constantly—things like transaction frequency, spending categories, and timing anomalies. For example, if we see a user who typically makes small purchases suddenly attempt a large transfer at 3 a.m., that’s a red flag. We also use a system that decodes messy transaction data into plain language, while a separate numerical engine looks for patterns in spending. Together, they build a dynamic risk profile that evolves with every action, allowing us to spot suspicious activity in real time, not after the money is already gone.
Given your work in markets across Russia, APAC, and the U.S., how have different regional fraud tactics shaped your overall strategy? Could you share a specific lesson from one market that proved invaluable in another?
Working in different markets taught me that data without business context is just noise. The fraud tactics shift, but the underlying principle of finding and exploiting weaknesses remains the same. A key lesson I learned in Asia, specifically while leading a team at Cashwagon, was the immense cost of bad data. We focused heavily on data quality, reducing issues by 35%. This principle is universal. In financial services, bad data leads to bad decisions, which costs real money. That foundational work on data integrity, like implementing a data warehouse that cut retrieval times by 40%, proved invaluable in the U.S. because it allowed us to build more accurate and faster fraud models from the start.
Business leaders often see data teams as a cost center, with different departments working in silos. How do you demonstrate the data team’s value as a strategic partner that connects marketing, risk, and product? What is the first step to bridging those divides?
That’s a perception I’ve fought my entire career. The problem is that departments optimize for their own little piece of the puzzle. Marketing celebrates a high number of installs, and the risk team celebrates a high number of rejections. But no one is asking the most important question: are the customers marketing is bringing in the same ones risk is turning away? This is where the data team becomes a strategic partner. The first step to bridging that divide is to become the team that sees the full picture and connects those dots. We have to show, with data, how a decision in one department impacts another. By providing that holistic view, you stop being a cost center and become the source of truth that guides the entire business strategy.
For a company new to data analytics, you advise against building a perfect, enterprise-wide system first. Could you outline the process for tackling one concrete problem to prove value and how you would measure and present that initial success to leadership?
Absolutely. Trying to build a perfect, all-encompassing infrastructure from day one is a recipe for disaster and frustration. The best approach is to find one team with one concrete, painful problem. For example, let’s say the finance team is struggling with a high volume of chargebacks. We would focus solely on that. The first step is to dig into their data, build a targeted model—like the liveness verification I mentioned—and deploy it. To measure success, we would track the chargeback loss rate before and after the change. A 35% reduction is a hard number that no one can argue with. You present that specific, measurable win to leadership. That success builds credibility and trust, which you then leverage to get the resources to tackle the next problem.
What is your forecast for the AI fraud detection arms race over the next five years?
The arms race is only going to accelerate. On one hand, generative AI is making it terrifyingly easy for criminals to create convincing deepfakes and synthetic identities. On the other hand, it’s giving us far more sophisticated tools to detect them. The companies that will win this fight are the ones building adaptive systems—models that don’t just follow static rules but learn from every single attack and update their defenses automatically. Analytics tools will evolve, too, moving beyond just flagging anomalies to understanding the legitimate business context behind them. Companies that are laying the groundwork now—cleaning their data, improving documentation, and building that foundational infrastructure—are going to have a massive head start. Everyone else will be left scrambling to catch up.