2026 · Fintech
G-Money
Advisory investigator workspace for a MoMo fraud pilot. Ranked alerts, explainable reasons, human-only enforcement.
0enforcement actions the engine can take on its own
- Result
- A working investigator workspace over eight de-identified Ghana MoMo alerts, with a shadow-pilot dashboard that measures explanation coverage and triage time.
- Built with
- Next.js, TypeScript, Tailwind CSS
Problem
Fraud investigators on a mobile-money pilot need to know which alerts to open first and why, without handing freeze or block decisions to a model.
What I built
- Rank-first queue with SLA, typology, and assignment state on every card
- Case views for reasons, evidence, and neighbourhood graph before a disposition
- Four-eyes confirmation on high-score alerts; freeze/block actions blocked on purpose
- Reconstructable audit of scores, assignments, and decisions
Outcome
A working investigator workspace over eight de-identified Ghana MoMo alerts, with a shadow-pilot dashboard that measures explanation coverage and triage time.
