About

I build full-stack web applications and data systems for workflows where information is incomplete, messy, or difficult to trust. Recent work spans a collection ledger for an undocumented catalogue, a no-login group-planning application, a self-hosted Python system, and a retrieval prototype built around public UH Mānoa records.

I am a computer science student at UH Mānoa, expected to graduate in December 2026. I work across product definition, data modelling, frontend and backend implementation, deployment, validation, and iteration.

The common thread is end-to-end ownership: understand the real workflow, build the complete path, and make the result clear enough to use and inspect.

01

Product and interaction design

Shape interfaces around real tasks, repeated edits, and clear feedback.

02

Data modelling and backend systems

Turn messy source material into explicit models, persistence, APIs, and workflow states.

03

Applied AI and retrieval

Build retrieval, citation, source-grounding, and evaluation paths without treating model output as automatic truth.

04

Deployment and validation

Treat hosting, checks, operational boundaries, and maintenance as part of the system.

How I judge the work

Clear workflows

Can someone understand the path through the system without fighting the interface?

Trustworthy behaviour

Does the system handle real states, edge cases, and messy source data without hiding the boundary?

Operational reality

Can it actually be deployed, hosted, observed, and maintained after the demo works?

Stakeholder contexts

Spire Hawaiʻi LLP

Forecasting workflow project with biweekly stakeholder meetings, progress demos, and feedback cycles around financial modelling, dashboards, and backend foundations.

CIO Council of Hawaiʻi

Funded education-focused project with recurring progress reports and a planned presentation to the Education Committee.

Independent products

Ahhgela Ledger

Collection ledger for an undocumented sticker catalogue with verified and possible entries, region navigation, and repeated editing workflows.

Hangout Planner

No-login planning app for availability, polls, notes, and group-facing results through shareable links.

SpineFrame

Self-hosted Python application with authenticated flows, persistent data, review/editing paths, and deployment ownership.

Mānoa Faculty Senate RAG

Research prototype built around public Faculty Senate archive records, with retrieval, citations, source-grounding, and evaluation limits.

VocabVerse

Japanese study app centred on songs, vocabulary, grammar, review flows, and persistent study data.