EcoCommons — national ecological modelling platform

EcoCommons is Australia's national platform for ecological and environmental modelling — species distribution, climate projection, biodiversity and biosecurity analysis — used by researchers and government across the country. I was its Technical Lead and Solutions Architect: I designed the platform and its infrastructure, and led the engineering team and partner institutions through to launch.

The hard part was never any single model. It was making national-scale, reproducible science work across nine partner organisations, four scientific disciplines, and a decade of incompatible legacy systems — and making domain experts productive on infrastructure they didn't build. EcoCommons re-engineered earlier research tools (BCCVL, ecocloud, CSDM) into one coherent, modern platform: a curated microservice “app store” of analytical workflows, an integrated data explorer wiring in national and global providers (ALA, TERN, GBIF, OBIS, IMOS), a cloud analysis sandbox with JupyterLab and virtual desktops, and reproducible-by-design environments so a published result can actually be re-run.

Two lessons from that work shape how I build AI systems now. First, reproducibility as a load-bearing requirement rather than a nicety — versioned environments, FAIR data, provenance on every result — which is exactly the discipline I now bring to evals and agent telemetry. Second, that the real engineering problem in a multi-stakeholder platform is making domain experts productive on systems they didn't design: give the expert a clean surface, hide the orchestration, keep the trust boundary honest. That turns out to be the same problem as agent design.

This is the national-scale, multi-institution, ship-it-to-launch credential beneath everything else on this site — the twenty-year foundation the frontier work is built on.