Politics can feel like it’s moving faster than anyone can process. The headlines refresh before the context does, and instability—at home and abroad—keeps compounding. In that kind of world, AI has turned into a steady companion: an information hub when you need grounding, and honestly, emotional support when the news won’t stop. It’s evolving faster than it used to, which makes learning with it—and learning to live with it—more important than ever.
The motivation is personal. I’ve seen enough, domestically and internationally, to know how tragic, draining, and difficult some realities can be to sit with. So the question becomes: what if the constructive side of tech power were used well? Countering erosion and polarization, widening awareness, and lowering the friction for people who want to engage. Exposure shapes what people notice and what they care about. The aim here isn’t “cool AI.” It’s using the tools well.
Design systems that can pivot across contexts. From research to production, build with modularity, transparency, and robustness baked in.
Applied science for public‑interest systems: measure, model, deliver.
I build monitoring‑grade data + AI systems for public‑interest research—multilingual ingestion, event extraction, geo/time analytics, and evaluation harnesses that stay attached in production. The aim is reliable signals that decision‑makers can trust.
Alongside applied projects, I chase foundational ML: transformer memory and long‑context modeling, retrieval‑grounded generation, grounded entity linking, and robust signal extraction under noise and drift. I like technical problems that shift capabilities across many systems.
Open to research labs, NGOs, policy teams, and industry partners building high‑stakes information systems. I can help with research design, data systems, modeling, and deployment. Low‑friction contact here.
A few papers and notes that reflect the research direction—kept brief and preview-only.
Event detection + interpretable early‑warning signals for shifts in protest, restriction, media pressure, and advocacy activity — designed for frequent refreshes and evidence traceability.
Tracks influence patterns across channels (diplomatic / economic / information / cyber) with consistent task definitions, careful normalization, and transparent aggregation.
Detects environmental shocks and social responses using multilingual event modeling + geo grounding, with downstream time‑series signals for monitoring and analysis.
Map‑first monitoring at subnational resolution, wired to two‑stage geo reconciliation (country → ADM1), standardized counts, and surge detection for rapid scanning and drill‑down.
These are representative, not exhaustive — the emphasis is on reusable infrastructure that transfers to new domains and stakeholders.
Previews only (first 2 pages).
If you’re exploring collaborations, I can share additional technical notes (evaluation harnesses, sampling QA playbooks, schema/patch patterns).
Typical problem spaces these systems support.
Low‑friction ways to reach me or explore artifacts.
Best for collaboration notes, dataset/tool questions, and applied AI opportunities.
Pipelines, experiments, and artifacts that support monitoring‑grade AI systems.
Where a lot of the public‑interest systems work lives.
Useful work still benefits from a little life.
About: // Taiwan‑born; culturally curious by default // I travel through food and the stories people tell through what they cook // Off the clock: dancing | handcrafts | graffiti‑inspired lettering // Creative practice keeps the work human and grounded //
