Taxonomy
A shared structure connecting topics, journeys, markets and research questions.
Research systems · AI-enabled research
Making research knowledge cumulative.
How I structured fragmented evidence into a traceable foundation for synthesis, retrieval and reuse.
43studies in the taxonomy
1shared evidence model
Humanreview at every claim boundary
The problem
A completed study is not the same as reusable knowledge.
Evidence existed across reports, presentations and working files. The problem was finding it, verifying it and determining whether it still answered the current question. I focused on the layer between storage and decision-making.
Architecture
Canonical records
Quotes, figures, sample context and provenance preserved as reusable evidence units.
Grounded assistant
Gemini and Claude workflows restricted to approved research evidence.
Reusable standards
Survey, documentation and QA patterns that reduced reinvention across studies.
Principles
AI as a retrieval layer, not an authority.
Every useful output needed a path back to its source.
A convincing answer was not accepted without support.
Market, sample, method and limitations remained attached.
Research judgment remained responsible for interpretation.
What this enabled
A foundation for research that accumulates instead of disappearing after delivery.
The work established a reusable evidence model and a controlled way to apply AI to synthesis, documentation and knowledge retrieval. No adoption or time-saved metric was documented, so the contribution is described as infrastructure rather than inflated impact.