An offline-first AI question synthesis platform for professional regulatory exams — a domain where a wrong answer costs a candidate their licence attempt.
Structured prompt chaining turns official curricula into draft questions, held in a staging quarantine that nothing escapes without a human signature.
A domain-expert review desk owns proofreading, difficulty calibration and statutory citation checks before anything is published.
CQRS dual-write keeps the production database and MeiliSearch in step, giving sub-20ms typo-tolerant search across statutes and exam points.
Prefer LinkedIn? Message me there, or see what I'm building on GitHub.