marcus.ng.俊豪

ABOUT

Theoretically grounded, production-minded. I build contestable AI systems for law, finance, and regulation. Domains where "AI said so" is never a defence. Most high-stakes AI failures aren't about model capability. They fail because reasoning cannot be audited, challenged, or defended when decisions are questioned. My work sits at the intersection of engineering, law, and accountability: designing infrastructure that makes AI reasoning explicit, traceable, and contestable rather than opaque. I'm the founder of Klavis (Regulatory intelligence for Vietnam) and co-founder of Veracia (Verification-first legal AI infrastructure for Singapore and APAC). At Veracia, we've validated our approach with mid-sized law firms and aligned to frameworks like MAS AIRG and the EU AI Act. Our signature approach: metadata-aware retrieval, citation-first architecture, and disputable reasoning graphs. We are applying research in computational argumentation (AAAI papers) and ongoing conversations with practitioners facing verification pressure in real matters.

EXPERIENCE

EDUCATION