Overview
Consilient operates in the heavily regulated anti-money laundering sector, recently launching a platform for detecting fraudulent transactions across multiple banks. Initially, the company trained ML models on sensitive banking data and returned them without exposing transaction information.
As a platform engineering consultancy, Concept to Cloud created software packages integrating with banking systems. This project showcases our data-centric application design services in action and developed the first cloud platform to manage and combine models from each bank. This advancement significantly improved cross-bank fraud detection efficiency.
Overcoming Development Hurdles
Banking environments impose strict restrictions. The challenge intensified due to varying data quality across institutions, making consistent model improvement difficult. The team maintained heightened vigilance regarding sensitive information while ensuring model reliability.
Key obstacles included:
- Data quality inconsistencies among different banking sources
- Privacy-protecting model training using representative data without raw data access
- Software compatibility across diverse banking systems
- Banking preference for container inspection rather than Docker deployment
“We had to think outside the box to create solutions that did not depend on the bank’s current software or infrastructure assumptions.”
A New Route to Success
The initial Apache Spark-based platform proved unsuitable, as it relied heavily on banking infrastructure during deployment and created operating system inconsistency challenges.
The team restructured using Hamilton pipeline engine and Polars, enabling:
- Efficient data processing
- Python-based feature writing instead of complex DBT SQL scripts
- C package compilation for enhanced security and portability
- Operation on any modern operating system without requiring a Python environment
Redefining Financial Crime Detection
The federated learning technology now manages datasets exceeding traditional memory limits while meeting modern financial crime detection requirements. The platform offers:
- Fully testable features for enhanced accuracy
- Compatibility across banking environments
- Easy installation on modern operating systems
- Faster feature iteration and deployment cycles
Achieving a Limitless Future
Concept to Cloud transformed Consilient’s fraud detection capabilities through advanced cloud architecture. Learn more about this project in our full Consilient case study, enabling secure management and integration of ML models from various banks while maintaining privacy and performance. The platform now integrates seamlessly with different banking systems. For more on our approach to secure cloud architectures, explore our cloud solutions services.
Ex-NASA engineer and cloud architect with over a decade of experience building scalable systems for startups and enterprises.
Work with Tom →