HousingGPT —
Intelligence for an industry, built by the industry.
We're researching a residential foundation model for Australian housing — trained on real development data, designed to answer the questions developers, builders, and planners actually ask.
Most AI is built by tech companies for consumers. HousingGPT is our research toward the opposite: a model built by the industry, for the industry — and improved by every project that runs through the group's systems. This isn't a chatbot. It's a foundation we're building toward.
Based on historical DA records for this Council, the most common delay drivers are stormwater management conditions, landscaping amendments, and late traffic referrals. Addressing these before lodgement typically shortens assessment.
Three components. One foundation.
The core architecture of our foundation-model research.
Housing Dataset
A growing body of first-party Australian residential development data — real projects, real outcomes, real timelines and costs. It grows with every project the group runs. (First-party and real-outcome — not forecasts.)
Vision & Multimodal Models
Research toward models that read site photos, drawings, and documents — turning unstructured construction reality into structured intelligence.
Knowledge Graph
Connecting sites, regulations, costs, and outcomes into a single queryable structure — so the model reasons about housing, not just text.
Built for the questions that matter.
- 1Zoning overlays are parsed into semantic constraints
- 2Model maps setbacks, height limits, and utility access
- 3Knowledge graph links site coordinates to historical local outcomes
- 4Generates probabilistic feasibility and material cost ranges
- Rapid site capability assessment
- Zoning constraint mapping automation
- Data-driven early feasibility estimation
- 1Weekly drone or site photos uploaded to the model
- 2Vision model identifies structural milestones completed
- 3Compares visual state against schedule timeline plans
- 4Flags delay risks and structural anomalies early
- Objectively tracked building milestones
- Early risk mitigation triggers
- Standardised progress records
From a team that publishes.
HousingGPT is being built by the same team behind three internationally accepted papers and a 2025 AUBEA Best Paper Award — with active university research partnerships. Our foundation-model research stands on a published track record, not hype.
Our foundation-model research is developed with active university research partnerships across computer vision and housing data science. Partnership details published as they are formalised.
Be first when it's ready.
HousingGPT is in active research and development. Join the waitlist to follow its progress, or partner with us to shape it.