I am a computational epidemiologist turned engineer, based in Auckland. My journey began in high-performance computing — simulating galaxy clusters on supercomputers — and has evolved through infectious disease modelling and peer-reviewed research into the cloud engineering and applied ML work I do today. The throughline is the same: rigorous thinking about complex systems, whether the system is a galaxy, an epidemic, or a production ML pipeline.
Today a university research clinic, a hospital neurology service, a health-tech company and a training institute each bring me in for the same thing: health data and models that hold up to scrutiny. I am the data scientist for the Neurogenetics Research Clinic at the University of Auckland's Centre for Brain Research, run the data behind the neurogenetics registry and biobank at Auckland City Hospital, lead engineering for a digital mental-health platform at Loffty, and lead the Data Science & AI programme at the Institute of Data. I am most engaged when the problem is genuinely hard and the stakes are real — health data, clinical informatics, any domain where the distance between data and a critical decision matters most. Whatever I am building, I try to bring the same standard: reproducible, honest, and built to last.
When I am not architecting systems or training models, you can usually find me exploring Auckland's craft beer scene, chasing the perfect cup of specialty coffee, or deep-diving into whatever complex problem has caught my attention that week.
Eight unbroken years of computational work, from supercomputers to hospital registries.
Research, clinical care, health tech and teaching: four teams that rely on the same skill.
Waipapa Taumata Rau | University of Auckland — Centre for Brain Research, Faculty of Medical and Health Sciences
Health New Zealand | Te Whatu Ora — Department of Neurology, Auckland City Hospital
Loffty
Institute of Data
Molecular Epidemiology and Public Health Laboratory
The University of Auckland
Massey University
Inter-University Centre for Astronomy and Astrophysics (IUCAA), India
Georgia Institute of Technology, Atlanta, USA
Data products, ML pipelines, and open-source research code
A visual, step-by-step guide to machine learning, from what a model is, through the data and maths intuition, to regression, gradient descent, overfitting and regularisation, and on to classic models such as decision trees, random forests, k-means and PCA, then neural networks, attention and transformers. Every number in the pictures is computed live, and every lesson shows the maths one click deeper.
Live decision tool that finds every branch of a chain within a chosen radius, adjusts Google ratings for review volume with Bayesian shrinkage, scores the latest reviews with an in-browser sentiment model, and ranks branches with a fully explained visit score. Runs at zero cost on Cloudflare Workers.
The product I build at Tasman Lab. A researcher adds their papers and gets an assistant that answers from them alone and shows the PDF page behind every answer. It reads the maths and tables, follows up on earlier questions, and keeps private papers private. The chat at the top of this page is my own assistant.
Most weather apps show one model. wx asks eight (ECMWF, GFS, ICON, UKMO, JMA, GEM, Météo-France, CMA) for the same place, headlines the median and shows how far they disagree. Works in a terminal via curl or in a browser.
End-to-end encrypted file sharing: files are encrypted in the browser, and the key travels after the # and never reaches the server. Each drop carries a Seal — the nine stars of Matariki at their real positions, turned by bytes of the key — so sender and recipient can confirm they hold the same drop.
Cross-source investigation into whether the Toyota Aqua's "most stolen car" status reflects genuine targeting or statistical illusion. Four data sources, 21,000+ theft records, time-series forecasting, and socioeconomic regression across 13 NZ regions.
A guided command-line tool for searching the OMIM API and exporting to Excel, CSV and JSON, written for clinicians without programming experience; and the auditable OMIM-to-PubMed completeness search behind a scoping review of hereditary neuropathy with auditory and vestibular involvement.
End-to-end ML evaluation product simulating 10,000-patient cohorts to benchmark AI screening tools. Modular pipeline architecture (config/, src/, notebooks/) with stratified accuracy metrics (sensitivity, specificity, AUC-ROC).
Multi-page data product with 7 interactive modules: KPI tracking, SPC charts, anomaly alerting, automated stakeholder reporting, and statistical hypothesis testing (Mann-Whitney U, Kruskal-Wallis, Chi-Square).
Production-grade data product: 34,000+ measurements processed through a four-gate ETL pipeline with audit logging, Power BI star-schema export, interactive Streamlit dashboard, and CI/CD via GitHub Actions with pytest validation.
Large-scale predictive model across 340 regions with multi-source data integration and parameter optimisation. Open-source, reproducible pipeline accompanying first-author publication.
The research assistant I build at Tasman Lab. Every way a researcher can share theirs is working on this page.
Bottom right of every page. Opens the same assistant in a panel.
For posters, slides and conference talks.
11 peer-reviewed articles • Google Scholar • ORCID
Journal of the Royal Society Interface, 19(187)
Visual summaries of selected research outputs and data products
I take on a small number of research and data engagements each year. If your team has hard health data and a decision riding on it, let's talk.
I build end-to-end data products — from ETL pipelines and predictive models to dashboards and APIs. 7+ years experience, PhD-qualified, with a track record of delivering measurable outcomes across cross-functional teams.
Send Me an EmailMy expertise spans computational modelling, AI evaluation, predictive analytics, and health informatics. I'm eager to collaborate on impactful, data-driven research across domains.
Propose a Collaboration