Pick a role. Only see the parts you care about.
Born in Taiwan, raised in Shenzhen, studying and working in Hong Kong.
Years across Taiwan, Shenzhen and Hong Kong gave me a fairly unusual view of things.
I mostly live in Nanshan, Shenzhen, work in Hong Kong, and go back to Taiwan when I can.
The chubby kid in the photos is me. I got leaner because I felt too fat and wanted to be healthier. And why not?
I think cutting weight, investing, and work are the same game: discipline, long-term, resilience, compounding, systems.
I study Data Science at HKU (graduating end of 2026), interned in analytics at HSBC CIB, and now build private AI systems inside a Hong Kong asset manager.
Outside work I make content on RedNote and Douyin as 小菜Nick. I like investing and asset allocation, cycling, basketball, training, and making things that actually matter — like using AI to solve problems for companies and friends, and creating value.
I used to be really fat
Not going back
time in the market > timing the market
Present
AI Transformation & Asset Management Analyst Intern
Building AI infrastructure and workflow automation for a Hong Kong asset manager while supporting private-credit investment processes.
2025
Corporate & Institutional Banking Data & Analytics Intern — Channel Analytics
Used analytics, machine learning and governed enterprise data to improve corporate-client onboarding and digital journeys.
01
An internal AI workspace that pulls together fragmented deal data, investment workflows and AI tools for investment teams.
It sits between messy source systems and the people who actually use them: documents, workflow state and model-assisted tools on one surface. No client or deal information is shown here.
02
A human-in-the-loop screening system that turns investment teasers into evidence-backed scorecards, so review is more consistent.
The agent extracts claims and asks for evidence. Judgment stays with the investment team. Consistency comes from structure, not from pretending a model can replace review.
03
Deployed Qwen on NVIDIA DGX Spark via vLLM to create a private inference layer for privacy-sensitive financial workflows.
Sensitive workflows stay on private infrastructure. This site only shows the architecture idea: local model, serving layer, and the systems that call it.
04
A Bloomberg and news-monitoring pipeline that automates recurring market research and portfolio monitoring.
Recurring research should not depend on someone remembering to check the same screens. The pipeline takes market and news signals and turns them into a repeatable brief.
A community of 150+ members, and 5+ industry events.
Cursor → GitHub → Vercel → nickts.ai
Next.js · TypeScript · Tailwind CSS · Motion
This is a personal homepage.
nicktsai1221@gmail.com
