Independent private product
Guardian
Personal health & performance platform
A private application for health records, training plans and research, with a clear separation between AI suggestions and user decisions.
React · TypeScript · Structured records · AI review workflows
The context
Long-running personal records need context and history. AI can help organise the material, but original notes, interpretations and accepted changes should remain distinguishable.
My contribution
I am building the application and its connected workflows for records, goals and research. The implementation keeps original material, extracted information, AI suggestions and user decisions separate.
What the work includes
- Journal entries preserve the original note alongside extracted data, hypotheses and review status.
- Goal changes begin as proposals. Accepted changes create a new version; a linked protocol has its own approval step.
- Research claims keep sources, uncertainty, versions and a record of review. A finding does not automatically change a personal plan.
Why it matters
These patterns also apply to business software: traceable information, controlled changes and a person making the final decision.
Have something similar in mind?