A.I.

I set out to prove this concept. My goal: build “BankAssist-SLM”, a specialized SLM capable of taking cryptic transaction strings (e.g., POS PUR 5942 MOONPAY.COM VALETTA MT) and accurately categorizing them, identifying the merchant, and flagging.

The "Cloud Tax" is currently a key impediment for many Enterprise AI strategies. Between latency, privacy risks, and exploding costs, scaling AI can feel like an uphill battle. But the solution isn't just "more cloud"—it's a smart Hybrid Architecture

The "AI tourism" in banking is officially over; the true differentiator now lies in the industrialization of intelligence.

The cloud-LLM paradigm is a strategic dead end—a high-latency, stateless, and unsustainable 'mainframe' model.

as AI permeates every facet of our lives, a fundamental gap persists: the absence of genuine subjective experience, or sentience.

A year ago, we launched the Backbase AI Center in Ho Chi Minh City with a bold vision: to build an elite team that would redefine the future of banking. Today, on our first anniversary, I'm incredibly proud to share this blog post.

Banking as we know it is a chore—a relic of a pre-AI era. We've moved from reactive dashboards to the promise of empathetic AI coaches, but the final evolution is far more radical: a completely invisible financial world.

We're all amazed by what AI can do, but behind the scenes lies a massive, hidden cost. From the energy it takes to train a single model to the shocking amount of water data centers consume, the environmental impact is real and severe.

One of the things I enjoy about business trips is the uninterrupted time they provide for reading. The longer the trip, the better. On my last one, I read Peter Godfrey-Smith’s Other Minds.