Architecting an advanced analytics platform for machine learning

Georgios Gkekas shares ING's advanced analytics journey to promote modern machine and deep learning techniques internally through a central, best-of-breed technical platform tailored for data science activities. The platform offers only the necessary automated tools to replace the tedious, repetitive, and error-prone steps in a typical data science pipeline.

As data increasingly becomes the new currency, the “triple A” concept of advanced analytics, automation, and AI has taken a central role within today’s enterprises. Modern corporations must make intelligent use of data—both their own and from third parties—if they want to ensure sustainable growth in the future and that they understand and are meeting customer needs. Currently, a myriad of data processing frameworks promise the “holy grail” of getting data insights into the executive suite. However, despite the abundance of frameworks, the effective integration of those frameworks into production environments, especially in highly regulated markets such as banking, still remains a huge challenge. The big data space has focused too much on covering the functional needs of the industry and has neglected the equally important nonfunctional ones, such as security, disaster recovery, regulatory compliance, and data backups, to name a few. Additionally, many companies struggle to find the right balance between the breadth of use cases covered by modern technologies and wide acceptance from CI/BI professionals. More often than not, executives, architects, and tech leads find themselves overwhelmed by the abundance of offerings and spend a substantial amount of time finding experts to use those technologies—not to mention how challenging it is to promote such a tech stack to a traditional in-house analytics team. Georgios Gkekas shares ING’s advanced analytics journey to promote modern machine and deep learning techniques internally through a central, best-of-breed technical platform, tailored for data science activities. This is easier said than done in a federated bank with worldwide branches and distributed and decentralized data sources. Georgios offers an overview of the platform, which facilitates data scientists’ daily work by offering only the necessary automated tools to replace the tedious, repetitive, and error-prone steps in a typical data science pipeline. He also demonstrates how to create a ubiquitous big data platform that