We from URA Research Group entered ATTACKER 2025: Are you an innovator? We’re your investors — Discovering FinTech & Digital Assets with one simple goal: build something Vietnamese banks would actually use. The sponsor roster featured heavyweight banks, so we aimed our research at a real pain point they all share, navigating Vietnam’s vast, fast-changing legal landscape. That focus became URA-xLaw, our AI legal Q&A platform for the State Bank of Vietnam context, and it carried Team URAx all the way to First Prize.
Inside Vietnamese banks, “please check the law” is a deceptively hard request. Vietnam’s legal landscape is sprawling, with hundreds of thousands of documents that evolve constantly—circulars, decisions, resolutions, and guiding notices that can shift the ground overnight. Teams spend hours searching, cross-referencing, and reconciling changes, only to repeat the process the next time an audit, product launch, or policy query arrives. We chose to attack that bottleneck head-on. That decision became URA-xLaw, our AI legal Q&A platform for banking compliance. Ask a natural-language question, and URA-xLaw returns a grounded answer with citations, effective dates, and the context needed to act. When the law changes, the guidance changes—automatically and visibly—so teams don’t have to wonder which clause still applies.
Under the hood, we curated a focused corpus of banking-relevant laws and normalized how those texts are parsed, linked, and updated. Our retrieval layer pulls the right passages, not just the right documents, so users see the clause that matters rather than a PDF haystack. The reasoning layer explains its steps and points to sources you can click and verify, because in compliance, explainability isn’t a nice-to-have; it’s the trust contract. On top of that, role-based access and audit trails make the tool enterprise-ready from day one.
In the finals, our team lead, Nguyễn Song Thiên Long, opened with a crisp narrative: a compliance officer under a ticking deadline and a single question threatening to stall a product launch. Then he put URA-xLaw to work—delivering pinpoint citations, change histories, and plain-English reasoning in real time. The judges leaned in; the room felt the relief of guidance you can both trust and defend. We wrapped to multiple “Shark”-style investors inviting us to take the product into real-world pilots.
We stood on a lot of shoulders to get here. We were guided by Prof. Quan Thanh Tho and Mr. Nguyen Ngoc Thai, whose feedback sharpened both the technical architecture and the business model. They pushed us to trade flashy features for reliable ones, to make traceability as central as accuracy, and to show our receipts every time the system spoke. Their fingerprints are on every good decision we made.
Team URAx is five people who believed utility beats novelty: Nguyễn Song Thiên Long, Võ Thị Như Quỳnh, Phan Quốc Khoa, Lê Ngọc Hùng Dũng, and Lê Thanh Duy. We argued over details, cut features we loved, and kept asking the same question: would a compliance officer use this every day? That single standard quietly shaped our roadmap and our demonstration.
For us, URA-xLaw is what happens when research meets a sharp constraint and a real customer. The sponsor list nudged us toward a genuine problem; the build forced us to earn trust with citations and updates; the pitch let us tell a story that belonged to the people doing the work. Winning Attacker 2025 is a milestone, not a finish line—and we’re excited for the future of AI.