MedTech & HealthTech

The sector is moving faster than the regulation

Two things changed in the last 18 months. Generative models crossed the threshold where they could help triage clinical signals without embarrassing themselves, and the TGA (along with its overseas cousins) started publishing actual guidance about how AI in health should be governed. The gap between what’s technically possible and what’s compliant is where most projects die. We build in that gap.

What we’ve shipped for health teams

• Rare-disease research platforms — small, messy datasets; consent lineage that holds up under audit; federated analytics so data never has to leave the custodian’s environment.
• Clinical-decision support tooling — always with a human in the loop, explicit confidence bands, and an override log a regulator can read.
• Patient-engagement flows — appointment, reminders, outcome capture; integrated with Cliniko, Halaxy, Nookal or your existing PAS.
• De-identification pipelines — Safe Harbor + expert determination, re-identification risk scoring, and the audit artefacts to prove it.

Compliance we actually know

Australian Privacy Act (APPs 1, 6, 11 in particular), My Health Records Act where relevant, AHPRA advertising rules for clinician-facing content, TGA guidance on SaMD (Software as a Medical Device), HIPAA-adjacent controls for projects that touch US data. We won’t write you a legal opinion — but we build to a standard that won’t embarrass your lawyer.

What we won’t do

We won’t build diagnostic tools that bypass clinician sign-off. We won’t ship unreviewed model outputs into patient-facing UIs. We won’t use patient data to train a general-purpose model without explicit, specific consent. These aren’t our preferences — they’re the table stakes of working in health.