Emotional-design systems, motion and design tokens โ the craft closest to the role I'm applying for. Self-contained, interactive galleries.
A 31-state emotional-design system for MewGuard โ every screen state (empty, loading, success, danger, recovery…) designed with one consistent emotional voice and character, so the app feels like a calm guardian at every moment.
The motion layer for MewGuard โ a gallery of hand-built Lottie animations for loading, feedback and celebratory moments, each tuned for timing, weight and personality.
A full design-system / design-tokens reference for Bopomofo Land โ colour, type, spacing and component tokens documented as a living style guide that keeps the product visually consistent at scale.
An emotional-design kit exploring mood-driven states, applied to Errand Pin โ a study in giving a deliberately minimal utility app warmth, character and a sense of care.
Projects where the AI or automation is the product โ systems and pipelines I designed and built.
Led the initiative that reshaped the Taiwan Chamber of Commerce in Denmark's daily operations around AI and automation. Worked directly with each process owner โ interviewing the finance lead to map her day-to-day workflow โ then redesigned those flows into automated, LLM-assisted pipelines. Covered the accounting reconciliation workflow and the member onboarding & verification workflow.
Cut accounting reconciliation from one person ร 3 days down to a single 2-minute, one-click run. Also ran a hands-on AI workshop that took 10 complete beginners from zero to shipping their own website.
A generator that turns a book into a beautiful, self-contained interactive course โ chapter breakdowns, plain-English explanations, embedded quizzes and visualisations. I designed and built the book โ course content-generation workflow: extracting structure from the source, shaping it into teachable modules, and driving a coding agent to assemble the output.
Course UI/design adapted from the open-source codebase-to-course design system (credit: Zara Zhang). The content pipeline is my own work.
A production retrieval-augmented assistant over an immigration-law corpus: answers questions with grounded source citations, backed by an evaluation harness that measures retrieval accuracy and citation correctness. Built from lived experience as an expat in Denmark.
Goal: reportable answer accuracy over a curated evaluation set, with every answer traceable to its source. Shipping soon.
Production iOS apps shipped solo โ product, full-stack and UX โ using AI-assisted development.
A cross-platform membership app for the Taiwan Chamber of Commerce in Denmark โ members sign in to browse curated partner perks, filter benefits by category, and register for community events. Published on the App Store.
Serving 40+ active members with ~100 monthly active users.
An iOS app that helps cat owners keep their pets safe from everyday hazards. Search any plant or food to check toxicity risk, browse curated cat-friendly plant guides, and manage cat profiles with a care timeline (meals, medication, play) โ all behind a subscription with a free trial.
A deliberately minimal location-based to-do app. Pin a task to a place, and a background geofence reminds you the moment you pass nearby โ so you handle errands while you're already on the way, with optional time reminders.
Five years building the scalable data platforms that analytics and AI run on.
Modernised the Finance department's reporting architecture: re-engineered legacy business logic into optimised Spark SQL and orchestrated scalable Databricks workflows for automated, reliable data delivery, working daily with analysts and PowerBI developers across departments.
As an IT consultant at Capgemini, built large-scale data pipelines for a major Danish telecommunications company โ translating complex business logic into scalable distributed processing on Databricks, with a heavy focus on PySpark and Spark SQL.
As an IT consultant at Capgemini, migrated a large manufacturing company's data platform from legacy systems to a modern Databricks lakehouse โ designing data-warehouse models and re-engineering ETL pipelines into optimised Spark for reliable, scalable analytics.
Achieved 10x+ performance gains in ELT pipelines by refactoring legacy code.