Data scientist Darwin · Northern Territory Available for work

I turn Territory data into decisions, not just dashboards.

Environmental intelligence, spatial analysis and full-stack delivery — taking open Territory data from a raw feed to something a council, a researcher or a traveller can actually act on.

Master of Data Science at Charles Darwin University. Treasury analyst turned data scientist. Environmental and biodiversity intelligence, spatial analysis, and fifteen shipped data projects, most of them built on Northern Territory data.

  • CDU Code Fair 2026 · Team of two
  • Optimisation
  • Language model triage
  • H3
  • CP-SAT
  • Python
  • MapLibre GL

Remote repairs ranked by need, with neighbouring communities sharing one trip— 9 in 10 urgent remote repairs fixed within 3 days instead of 62, across a synthetic year.

  • Machine learning
  • Spatial
  • Forecasting
  • Gradient boosting
  • LSTM
  • Earth Engine
  • Python

Every 5 km of the Territory classified by when it burns, 2000 to 2025— and a 2026 forecast: 51% expected to burn, most of it after July.

  • CDU Code Fair 2026 · Team of two
  • Network analysis
  • Open data
  • ACMA register
  • Python

A coverage map shows a signal, not what it depends on— 18 of 23 places on a radio chain rely on at least one relay with no other licensed path.

NT·04 TerraIQ Live A Territory-wide weather memory built from 44 BOM stations, holding what the public feed drops after 72 hours. It collects and checks itself on a schedule, with no server to run. Open → NT·05 Conflux Live Air and sea, compared directly every three hours. Two anomaly detectors that shared their inputs still disagreed, and the dashboard says so. Open → NT·06 Avian Observatory Research Pointed at Territory audio, BirdNET confidently answers Bicknell’s Thrush — a bird of the north-eastern US. A regional model for 449 species names the right bird 80.6% of the time on 537 recordings from 2026, against 65.7% for BirdNET. Open → NT·07 BushMetrics Live 24.6% of the Territory is protected, and Traditional Owners manage or co-manage 92.6% of it. 8 of 25 bioregions meet the 30% target; the gaps run through the pastoral middle, down to 0.7% on the Sturt Plateau. Open → ML·08 RainSignal AU PRT565 · Team of three Live Bureau of Meteorology observations run through frozen rainfall models trained on 267,317 station-days. An observation and an estimate are never shown as the same thing, and stale data is labelled stale. Open → NT·09 RoadState Live The Territory's real road problem is seasonal access, not congestion: 78% of traffic sits in Darwin, and 66% of the year's closures begin in the wet. Open → NT·10 NT Crime Intelligence Live Switching from raw counts to per capita completely reorders which communities are most affected. How you frame the data decides who gets served by it. Open → RE·11 Biodiversity Surrogate Validation Research A pre-registered, multi-catchment test of open spatial surrogates against expert assessment. Vegetation-type rarity won, but only modestly — these are screens, not substitutes. Repo → RE·12 NT Conservation Exposure Research A transparent, open-data conservation exposure index for the whole Territory, tested against expert biodiversity assessment of the Roper River. Agreement was weak, and it came from land convertibility, not biodiversity. Repo → CIV·13 ArmaWatch Delivery A mobile-friendly rebuild of Wage Peace's Australian weapons-industry map — 127 sites, clustered mapping, filters, live summaries and CSV export for research and journalism. Open → ML·14 Gradaroo Live An AI-assisted graduate job finder for Australia — 40 universities, live Adzuna listings, Gemini resume matching. Curated lists and approximations are labelled distinctly; nothing is scraped. Open → DA·15 Australian Charities Analytics PRT564 Descriptive, predictive and diagnostic analysis on the ACNC Charity Register — research-question design through to a finished visualisation pipeline. Repo →
ML·16 Student Central RAG Early stage A retrieval-augmented assistant answering student questions from official CDU content with citations — and drafting an escalation email when it cannot confidently answer. In progress

How I work

I start with the question, not the dataset.

Open data is easy to download and hard to make mean anything. Before I write a query I want to know what decision is waiting on the answer.

I ship the tool, not the notebook.

Analysis that lives in a notebook dies with the kernel. Every project here ends as something deployed and reachable.

I say what the data doesn't support.

The surrogate validation found the surrogates only modestly useful. Conflux expected two detectors to agree and they didn't.

I build it so it keeps running without me.

Scheduled jobs, no servers to babysit, a plain-English reading on every chart. If you have to ring me to interpret it, I haven't finished.

For teams

Data science roles

For teams who need someone who can do the analysis and ship the thing around it — the model, the pipeline, the map, the deploy. Master of Data Science coursework completed at CDU, with a Treasury analyst's instinct for what a number is actually going to be used for.

For researchers, councils & not-for-profits

Project collaborations

For people sitting on data they can't yet use. Biodiversity records, sensor feeds, service statistics, survey work — I'll help turn it into something reproducible and public-facing, the way ArmaWatch and the surrogate validation were done.

All articles →