Practitioner-focused whitepapers on real-time Anti Money Laundering and Fraud Detection — written by the developer of Jube from direct experience in the field.
New whitepapers are published periodically. Check back here, or subscribe to be notified by email when the next one drops.
July 2026: Labrador-Level Advanced Analytics: Fraud and Risk Models from Someone Who Never Quite Made It to School
A plain-speaking tour of the statistical models behind fraud and risk scoring, written in retort to “it is just a bunch of if statements”. The paper dissects rules into Abstraction and Activation, makes the case for a single well calibrated quantitative score — the thermostat by which risk appetite is turned — and walks through Logistic Regression, Neural Networks, and decision trees before arriving at the author’s weapon of choice, Bayesian Networks via bnlearn in R. Along the way it covers structure learning, bootstrap arc strength as regulator-ready validation, and the practicalities of deploying models as Plumber-wrapped HTTP services, all anchored to the generally accepted practices of Basel model governance rather than the bravado of AI. Includes a required reading list spanning Baesens, Pearl, Neil, and Scutari, and concedes — in the end — that the accusation is half true. Written for practitioners without a rigorous academic background who nonetheless need defensible analytical work product, by one of their own.
June 2026: No Big Rubber Shoes: A Jaded Practitioner’s Guide to Building on Jube
A practical guide for System Integrators embarking on a Jube implementation for real-time Anti Money Laundering and Fraud Detection. The paper argues for a disciplined, iterative approach over big-bang delivery, grounding the engagement firmly in regulatory satisfaction as the singular objective. It covers integration realities, the dangers of freestyling without Run Books, the importance of analytical methodology over template rules, and how SIs can build proprietary IP within Jube’s extensibility framework while remaining compliant with its AGPLv3 licence. Candid, experience-led, and written for practitioners who want to build a practice rather than merely survive a project.