Our approach
In health technology, how you build matters as much as what you build. Our approach is defined by safety, transparency, and respect for the people we serve.
In health, how you build is as important as what you build
Most health technology fails not because of bad algorithms, but because of bad assumptions about who the user is, what they need, and how much risk they can absorb.
We start from a different place. We assume that health software carries real consequences, that the people using it are often in vulnerable situations, and that trust, once broken, is almost impossible to rebuild.
That belief shapes every decision we make: what we build, how we build it, and what we choose not to build.
Structural safety, not behavioural safety
Many AI systems rely on behavioural guardrails: instructions that tell a model what not to do. In health, that's not enough. A prompt telling an AI to "not give medical advice" is a policy, not a safeguard.
We build structural safety into our architecture. That means:
- Constraining what the system can access and output at the infrastructure level
- Separating clinical content from AI-generated summaries with clear provenance
- Designing for auditability, so every recommendation has a traceable source
- Testing for failure modes before testing for features
We'd rather ship less and ship safely than move fast and hope nothing breaks.
Built to meet Australian regulatory standards
IEC 62304
Medical device software lifecycle processes, which define how we plan, develop, and maintain our software.
ISO 14971
Risk management for medical devices: a systematic approach to identifying, evaluating, and controlling risk.
ISO 13485
Quality management systems for medical devices, ensuring consistent design, development, and delivery processes.
TGA compliance
Australia's Therapeutic Goods Administration pathway for software as a medical device (SaMD), our target regulatory framework.
We're building toward these standards from day one, not retrofitting compliance after the product is built.
Our position on AI in health
Explainability
If our system surfaces a recommendation or summary, the user must be able to understand why, and trace it back to the source. No black boxes.
Human authority
AI supports human decisions. It never replaces clinical judgement, and it never acts autonomously on health-critical actions.
Bias awareness
Health AI trained on biased data will produce biased outcomes. We commit to ongoing auditing, diverse training data, and transparent reporting of known limitations.
Privacy
Health data is among the most sensitive data that exists. We treat it accordingly, with encryption, access controls, minimal data collection, and no secondary use.