Evidence stays visible.
Claims need a trail.
I prefer systems and writing where sources, assumptions, scope, and decisions stay inspectable instead of being buried behind polished language.
About
I build evidence-aware AI systems for high-stakes work.
I am a doctoral researcher at National Yang Ming Chiao Tung University, working across trustworthy AI, speech intelligence, clinical workflow support, cybersecurity, agent governance, and AI systems engineering.
My work is shaped by research training and earlier cybercrime investigation experience. I care about systems where claims stay grounded, boundaries are visible, and humans can review what the AI did before it becomes an operational decision.
Working model
Boundaries are design controls.
Clarity builds trust.
Clinical, security, privacy, and publication boundaries are part of the architecture. Narrow scope is useful when it makes review and validation stronger.
Build the smallest honest demo.
Proof before expansion.
The useful first artifact is often a narrow workflow that makes one claim testable: a staff-review summary, an evidence packet, a voice turn, or a validation gate.
Translate across audiences.
Research should become usable.
I care about turning technical research into reviewer-facing manuscripts, demos, teaching packets, and operational checklists without losing claim discipline.
Where the work points
The common thread is practical trust: speech systems that preserve decision evidence, clinical demos that stay synthetic and staff-review oriented, governance work that preserves reconstructability, and teaching materials that help teams deploy AI with infrastructure, security, evaluation, and delivery discipline.