I build and verify the knowledge bases behind production conversational AI.
Over the past year: 1,500+ knowledge base entries across 12 client deployments in 7 industries and 8 markets, plus the test suites that verify them.
My work centres on one question: can this answer be traced back to a source?
In practice. Control-based test methodology that separates genuine model failure from grading error. Mechanical validators that refuse to ship a file when a rule is violated. Retrieval diagnosis, where keywords determine eligibility and content determines ranking. Objective measurement in place of subjective argument.
Trained in archaeology, which is source criticism as a discipline: date every source, rank its authority, and never let inference pass as evidence.
Contact: araihan.rizkiwork@gmail.com