Case Study

Detect Money Laundering patterns

Detect Money Laundering patterns

Finding the bad guys among billions of transactions




Banks around the world need to fight money laundering to keep people safe and to protect themselves. Money laundering helps criminals hide their illegal activities, and it's a big problem for everyone. If banks don’t work hard to spot and stop money laundering, they can get into serious trouble. For example, in 2012, the big bank HSBC had to pay a huge fine of $1.9 billion because it didn’t do enough to stop money laundering. And in 2020, Westpac, a big bank in Australia, had to pay an even bigger fine of 1.3 billion Australian dollars for the same reason. These huge fines show how important it is for banks to follow the rules and work hard to stop money laundering. Doing this helps keep everyone safe and protects the banks too.




Banks face multiple challenges in the fight against money laundering, especially when using outdated legacy systems and dealing with well-funded and adaptive criminals.

  • Legacy Systems:

    • Slow and inflexible systems can miss suspicious activities.

    • Data about customers and transactions are spread around multiple systems, managed by different teams.

  • Well-Funded Adversaries:

    • Criminals invest heavily in finding new ways to bypass bank defenses.

    • Constant adaptation of tactics makes detection and prevention harder.

  • Costly and Time-Consuming Integration:

    • Upgrading to advanced systems is expensive and slow.

    • Requires staff training and process adaptation.

  • Risk of Non-Compliance:

    • Banks risk massive fines and damage to their reputation.

    • Failure to update systems and strategies leaves banks vulnerable to money laundering activities.




To tackle these substantial challenges, a two-pronged approach was employed, ensuring not only compliance with global standards but also a proactive stance against money laundering activities.

  • Utilization of Modern Data Platform – Foundry:

    • Leveraging Foundry’s advanced data integration and analysis capabilities for real-time insight.

    • Enhances the speed and accuracy of detecting suspicious activities by processing large volumes of data efficiently.

  • Adoption of Machine Learning Models:

    • Machine Learning (ML) models outperform traditional rule-based systems.

    • ML models can learn and adapt to new patterns of money laundering that criminals employ.

    • Enables faster and more accurate detection, reducing the risk of non-compliance and associated fines.




Utilizing Palantir Foundry, we effectively streamlined data operations:

  • Ingesting Regularly Data: The platform enabled seamless data ingestion from various global systems, streamlined creation of ETL pipelines for optimal data transformation and storage, and consistently ensured data quality and accuracy with built-in health checks.

  • Machine Learning Modeling: building and training the model is only part of the effort.

    • Feature Engineering: Specific features, such as anomalous large transactions and the in/out flow patterns, were meticulously analyzed to enhance the detection model. Additional assessments included examining frequent short-term deposits, multiple small transactions, and inconsistent international transactions, contributing to a comprehensive and robust detection system.

    • Experimentation with Diverse Machine Learning Models: Various machine learning models were tested to ascertain the most efficient and reliable one for identifying potential money laundering activities. This extensive experimentation ensured heightened accuracy and reliability in the detection process.

    • Prioritization of Model Explainability and Bias Checking: The use of SHAP values effectively explained the predictions of the machine learning model, fostering transparency and accountability. A rigorous bias checking process was also implemented, ensuring the delivery of fair and unbiased solutions while enhancing the robustness and reliability of the money laundering detection system.




The use of Machine Learning has various advantages when compared to traditional rule-base systems. The approach described above allowed the bank to get important improvements:

  • Enhanced Detection: The bank has significantly improved its ability to identify suspicious transactions. Detailed analysis and advanced machine learning models heightened the precision and reliability of the bank’s detection system, ensuring a more thorough investigation of potential money laundering activities.

  • Streamlined Data Management: By utilizing Foundry, the bank ensured smooth and efficient data management, enhancing the consistency and accuracy of its data. This improvement allowed for more timely and effective analysis of transaction patterns, further strengthening the detection capabilities.

  • Transparent and Equitable Model: A commitment to transparency and fairness in the bank's operations was prioritized. Utilizing SHAP values and conducting meticulous bias checks ensured that the anti-money laundering model operates transparently and equitably, ensuring fairness and accountability in its functionality.

  • Boosted Compliance and Security: The bank has fortified its adherence to global standards, enhancing its overall financial security and reducing the risk of non-compliance and associated fines.

These comprehensive improvements underscore the bank's unwavering commitment to providing a secure and compliant banking environment, effectively safeguarding both its operations and global financial integrity.

Machine Learning is an umbrella term representing various powerful techniques going from Linear Regression to Deep Learning models, like LLMs (ChatGPT-technology). It is a great tool to solve a variety of problems in business, absolutely not restricted to banking, it has been used with success in many industries.

Contact us and let's chat about your specific needs.

Contact Us