Anti-Money Laundering

A smart AML anomaly detection system for a London based global payments platform that wanted to identify suspicious activity across multiple payers and beneficiaries.

Anti-Money Laundering

The Opportunity

Our client handles thousands of cross-border transactions a day, and is responsible for the compliance and risk associated with these payments.

We were required to build an unsupervised anomaly detection system that could identify novel examples of suspicious behaviour on the platform.

What we did

✔︎ Initial exploration of suspicious activity on the platform
✔︎ Ingestion of transactional, payer and beneficiary data into graph database (Neo4J)
✔︎ Algorithmic detection of anomalous patterns in the graph DB
✔︎ Intelligent scoring system to produce interpretable results
✔︎ White paper outlining the techniques applied and novel examples of anomalous behaviour

Anti-Money Laundering


distinct behavioural patterns are calculated for each payer to generate their risk score.

How we did it

In the opening phase of the project, the priority was to explore and understand the available data, and to collaborate with internal stakeholders to design a system that would produce useful, actionable insights. Through regular workshops we designed the format and regularity of the scoring, and how our output would fit in with the analysts’ workflow.

Without labelled data, we needed to design a system that could identify anomalies in an unsupervised manner. Of equal importance was the ability of the system to be interpretable, so typical black-box approaches to anomaly detection would not be suitable here.

We quickly established that a graph database solution using Python and Neo4J was the optimal way to tackle the problem, as it is perfectly suited to running network algorithms and visualising the anomalous webs of behaviour across payers.

We iterated on our outputs through further feedback sessions with internal stakeholders and analysts in order to deliver a system that could highlight suspicious behavioural patterns across a range of factors, such as payment digit distribution and cyclic links between entities in the graph.


We asked ADSP to build an anti-money laundering (AML) model that could identify new kinds of suspicious activity on the platform. They delivered exactly what we needed and more. They had a fresh perspective on long-held challenges, and clearly demonstrated the value of using graph database technology to tackle money laundering. Their creative approach enabled us to quickly prove the value in augmenting our current monitoring systems with pattern detection.

Jan Philippaerts, VP Compliance Operations


Download our case study


Start a conversation

Take the first step by speaking with one of our data experts today.

Predictive Maintenance
Predicting Optimal Car Configurations