Designing a real-time machine learning platform for a large bank
Overview
Industry
Location
Financial services
Poland
Technology Used
Ansible
Apache Airflow
Apache HBase
Apache Hive
Apache Kafka
Apache Ranger
Apache Spark
Java
Python
Rundeck
Schema Registry
Show more
Project overview
Designing a platform to integrate AI and ML algorithms in a highly flexible and configurable manner, so that platform operators are able to incorporate the provided models into pipelines through configuration instead of custom coding. The introduced configuration included defining input and output data models, specifying communication channels for data transfer, enabling feature store management and retrieval, and applying data transformation logic before feeding the data into the model. Additionally, the platform supported the parsing of BIK (Credit Bureau Info – a data source containing information provided by banks, state banks and leasing companies in Poland), during the customer information preprocessing phase.
Client background
One of Poland's leading banks, our client provides services to individual and business customers and combines traditional banking principles with innovative solutions.
Goals
Managing high processing volumes each second, ensuring batch latency is under 5 seconds, and implementing complex transformation logic – while accelerating the introduction of innovations and guaranteeing 24/7 availability.
Results
Easier integration
Unifying the configuration process and production deployment, making integration across the organization more accessible and scalable.
Faster loan processing
Creating a platform to enable faster consumer loan processing in bank branches and online channels.
Solution for enhanced banking
Introducing a solution that allows for high reliability and auditability.
Streamlining the messaging process
Increasing the efficiency of processing incoming messages from BIK.
We'd love to hear from you!
Fill out the form - we'll get back to you as soon as possible
