Deploying AI for a Greener Future: Insights from the Field

Data Quality Over Volume

Rather than accumulating massive datasets, Hubble emphasises rigorous preprocessing and standardisation. The team generated 500,000+ datapoints in partnership with WSAA to analyse shading products across varied building conditions and climates.

Deliberately synthesised datasets ensure coverage across performance extremes — from poorly performing to highly efficient buildings. This variance prevents bias toward average outcomes and strengthens model reliability.

Regional Customisation Matters

Models require localisation to specific climates, construction practices, and regulations. A Victorian housing model won't transfer effectively to Queensland or arid zones without adjustment.

Transparency Builds Adoption

Clear explainability through prediction intervals, transparent assumptions, and scoring frameworks significantly increases stakeholder confidence and practical implementation.

Behavioural uncertainty is critical — occupant behaviour introduces variability that's harder to quantify than building physics. Incorporating behavioural factors improves real-world accuracy versus laboratory conditions.

Impact Over Metrics

Real success requires tangible outcomes — accelerated retrofits, integrated sustainability lending, informed homeowner decisions — not just statistical accuracy measures.

Recommended Experts

For those interested in this space, we recommend following:

• Rafa Felix, PhD — building physics and ML applications
• Dr. Clayton Miller — digital twins and data science
• Sara Behdad, PhD — sustainable design and circular economy
• Bahar Dadkhah, PhD — AI-driven energy modelling