| Ask SafeFuture Lab | A public agent whose answers are machine-checked against sources; it refuses when the check fails. Open → |
| Road Risk Explainer | AI assistant for AusRAP and network-risk concepts. Open → |
| Safe System Assessment assistant | AI assistant grounded in Safe System Assessment practice. Open → |
| Verified road-toll report | Queryable road-toll report with deterministic checks on every number. Open → |
| Crash-likelihood modelling | Machine-learning crash likelihood across the Victorian network on open data; a full-chain strategic road-trauma model prototype. |
| LLM crash-coding benchmark | Blind test of whether language models can assign the official crash-type code from a police narrative. |
| Offline de-identification | Local language models that strip personal details from crash narratives without data leaving the machine. |
| AI vision at scale | Reading thousands of scanned engineering sheets, with each model's misses measured. |