Collection with provenance.
Keep sources, permissions, and missing information visible. Responsible dataset collection begins before any model sees the data.
Series 2 · Developing in parallel
SignalMatters by Sharedstep
Our Total Health direction brings together health-data collection, datasets, and evaluation using existing AI models. Privacy, security, and clear permissions are central design requirements.
In developmentPreserve where information came from and what may be missing.
Keep care sharing, research evaluation, and model training distinct.
Compare compatible model outputs and document the limits.
Illustrative workflow · Development direction
Total Health · SignalMatters
Series 2 work continues in parallel at a lower priority while EyeOps and Model Hub lead Series 1.
Keep sources, permissions, and missing information visible. Responsible dataset collection begins before any model sees the data.
Track versions, intended uses, and access requirements. Permission for one use must not silently become permission for another.
Use compatible current AI models and supplied outputs to examine evidence, compare results, and record limitations. A model’s output is not a complete health assessment.
Privacy & security by design
These are requirements for the product we’re building. They must be implemented and verified for the intended deployment before handling real patient data.
Explain what is collected, why it is needed, and who may use it.
Verify access controls, protected storage, and recovery procedures for the deployment.
Make sharing, portability, withdrawal, and retention limits understandable.
What to know
Total Health describes the broader health direction within SignalMatters by Sharedstep. This work connects collection, datasets, and model evaluation.
The current direction emphasizes evaluation using existing compatible AI models. Any future training use needs its own data rights, permissions, scope, and evaluation.
No. This is an informational website with no health-record upload or patient account service. Patient-data deployment protections have not been established.
No comprehensive health score, diagnostic service, or clinical validity is claimed. Current work is a research and evaluation foundation.