DenizBank supports data scientists with its pioneering Data Warehouse Project

DenizBank, Intertech and Microsoft have teamed up for a Data Warehouse Project that provides Data Science departments with a flexible, research and production-oriented, cost-effective and scalable ecosystem and sets a precedent model for Türkiye as well as most markets in Middle East and Africa.

The project, which responds to the data science related critical-needs in banking and finance, has its R&D and cloud activities powered by Microsoft. It transfers sensitive data to the cloud-based environment by passing through certain mechanisms in compliance with the security criteria approved by the BRSA (Banking Regulation and Supervision Agency). The project provides a wide range of modelling capabilities from reporting to real-time online large data solutions and has increased the coefficient of productivity by solving problems at the source and creating new at source data structures. On the other hand, it is deemed an exemplary model that takes advantage of the power and speed of the cloud while remaining true to the principles of data security and privacy.

Time saving with innovative technologies

The project focuses on performance and data quality. Desired modelling can be performed in large data environments created with robust and real-time data, with statutory reporting provided in the right infrastructure. The project actively utilizes Always on technology which focuses on high accessibility and disaster recovery solutions and works simultaneously to generate specific data at the source, work on reports, and make improvements in the big data field.

An exciting environment with a wide range of opportunities for data scientists

Responding to the analytical and high-level in-house data needs of its 80-strong data science team through this collaboration, DenizBank also ensured a more advantageous cost structure and a higher level of speed and innovation for its big data operations. 

Data is stored on DenizBank servers through the Azure Machine Learning service infrastructure enabled environment that is compatible with LPPD (Law on Protection of Personal Data) and GDRP. As such, it is an exemplar model for the rest of the industry. It is only the authorised bit of data that is transferred to the Azure cloud environment and all transfers are carried out on an encryption-only basis. Data science teams have the chance to expand their Private Cloud resources as and when needed. Efforts are further enhanced by the fact that Azure provides various versions, latest updates, and tools from a single source point.  By moving data to the cloud in a daily scheduled infrastructure, DenizBank has thus created its own hybrid structure in data science.

Comprehensive and effective management of all applications with Azure

The project kicked off in November 2020. The first two months were used to identify the main sections after which efforts picked up speed with works concluding for implementation from start to finish in three months. At present, teams use the system for any needs including basic banking and financial applications by marking the data regarding the models they want to work with and all the while in accordance with relevant legislation. System makes it possible to work models with Azure Machine Learning Operations (MLOps) capabilities, to store results directly in the cloud, and to scale in cloud different data sets regardless of resources or effort. Much of the burden in terms of workload and cost, such as hardware, security, installation, and authorization are eliminated. In addition, the models can be delivered as services to DenizBank servers, and the required services can be delivered via the Azure Kubernetes Service infrastructure.

It provides effective management for synchronizing and updating all applications, while working with up-to-date and development-oriented resources based on the infrastructure for Machine Learning libraries and supported external libraries. Resources can be up and downsized in accordance with the objective of a flexible environment for model generation and with access to aged data, teams have the flexibility and speed they need.  

Rigorous work, strong collaboration between data science and IT teams, a pioneering level of work completed in full compliance with regulations and the capabilities provided by the cloud inspire people with the possibilities to accelerate model development and deployment through monitoring, verification, and management of machine learning models.