I build and evaluate machine learning and AI systems, and work with data through analysis and visualisation. I also build websites, web apps, and mobile apps.
I'm a data scientist in Sydney with a Master of Data Science from Macquarie University. My work runs from data analytics and visualisation, through statistical modelling on large datasets, to building and evaluating machine learning and AI systems. I also design and build websites, web apps, and mobile apps.
I came to data science from mechanical engineering, which is where the comfort with maths and the habit of measuring things carefully come from. I like problems where the data is messy and the answer matters, and I tend to build something to solve them rather than leave them on a slide.
A commit history of the education and work that got me here, newest first.
The roles behind the timeline, in a little more detail.
Python
R
PyTorch
JavaScript
React
SQL
Git
HTML5
CSS3
Sass
Supabase
Cloudflare
Figma
BlenderA mix of research, machine learning, and things I have built end to end.
My master's major project (Team, graded 82 / Distinction) on getting language models to disambiguate clinical abbreviations. I owned Task 1: framing it, cleaning the CASI dataset of ~26,500 clinical notes, and building the BERT and BioBERT baseline the team measured against.
Built at a hackathon for Global Recordings Network, who catalogue audio in thousands of rare languages. Full fine-tuning of Meta's 4,000-language speech model needed ~370GB RAM and 7 hours per epoch, so I trained a light classifier head on embeddings from the frozen encoder instead.
A full modelling project in R on Fannie Mae loan data: 420,000+ loans and 75 raw fields. I engineered a binary default target, then cut the fields to a 13-predictor logistic regression using backward stepwise selection on AIC, handling multicollinearity along the way. Delivered as a reproducible Quarto report.
A pickleball-gear storefront I designed and built end to end. It's a headless setup built with Lovable, wired to Shopify's Storefront API and hosted on Cloudflare, with the brand, product system, and architecture all mine.
A complete devotional mobile app I built as sole developer for a ministry, in React Native (Expo) with a Supabase backend. I handled all the engineering and the full build pipeline through to a tested Android build; the content comes from the ministry.
An academic survey on Multi-Instance Learning, a weakly-supervised approach where models learn from group-level labels instead of labelled individual points. It maps the main families of methods and reviews published object-detection work including WSDDN and MIL-SERBoost.
Open to data science and analytics roles, and happy to talk through any of the work above.