- Period
- 2025—NOW
- Surface
- WEB + MOBILE
- Role
- PRODUCT / ENGINEERING
- State
- Building
Raivora began as a personal need and expanded into a larger product question: can a wellness system accept how people really describe food without surrendering accuracy, privacy, or control?
The interface stays conversational, but conversation is not treated as truth. Structured input crosses a deliberate boundary; backend data and services calculate the answer, authorise the action, and ask for confirmation before a write becomes real.
How can natural language feel forgiving while the underlying health record remains precise, inspectable, and under the user’s control?
Separate interpretation from calculation. The assistant may understand intent and shape structured input; authoritative services validate quantities, calculate results, enforce access, and preserve confirmation boundaries.
Evidence of the work
- 01Messy multilingual meal and recipe input
- 02Explicit confirmation before consequential writes
- 03Backend-owned nutrition calculations
- 04Web and native client surfaces
Field principles
Helpful does not mean presumptuous.
The model proposes; the system verifies.
Private context should have an explicit boundary.
