Speedcubing timer and utility app. Built end-to-end with SwiftUI and UIKit. App written fully in Swift with some bridged C++ statistics functions.

Social enterprise platform for teenagers. Worked closely with android development team to build cohesive experience throughout platforms. Built with Swift, UIKit and assisted backend development with Prisma, Typescript and Next.js.

End-to-end web app that unites Aotearoa's eight local non-profit speedcubing clubs, written in SvelteKit. Backend built with Prisma, Kysely & PostgreSQL, and hosted on Oracle Cloud.

A portal interface for any server—infinitely customisable and scalable. Built with Svelte.

Participated with team from Machine Learning @ Berkeley. Finalist and highest ranked team in qualifying round with 0.08174 MAE. 2nd Place at in-person finals at Harvard! Built custom ResNet based model to predict solar panel PV generation, with data engineering and custom baked dataset.

Computer Vision project segmenting a Rubik's Cube and detecting the state through a video stream, producing a solution to solve the scrambled state. Started off as a classical computer vision project in Cal Hacks 10.0, built with OpenCV (C++) and bridged to Swift as an iOS application. Continued as part of the Machine Learning @ Berkeley's NMEP final project, used DINO, LangSAM (Langauge based Segment Anything Model) and OpenCV.

Simple utility iOS & Mac Catalyst app for generating shades and tints of a colour. Vectorised utilities for working with colour spaces and converting between commonly used colour representations.

POS ordering system with pre-order integration, trend analysis, stocking and finance management. Built with Vanilla JS. Python Flask API deployed with Azure & MS SQL Server, with automatic CI/CD.