With the Eastwind product, the operator improved the work of the big data analytics department, which helped it speed up the time-to-market of analytical ML models. MegaFon Tajikistan is a leading telecom company and the primary provider of digital services in the Republic of Tajikistan. It was the first in the country to launch 3G, 4G/LTE, and 5G cellular networks and holds a leading position in terms of market share by subscribers.

From a closed module to transparent work with data

For over six years, the operator has been using Eastwind's CVM solution, Mobile Marketing Platform (MMP), to personalize marketing, increase the conversion of advertising campaigns, and foster subscriber loyalty. Previously, this work was based on customer data analysis within MMP using its built-in analytical module.

When MegaFon Tajikistan created new independent departments, such as business analysis, CVM, and Data Science, Eastwind proposed expanding the MMP solution with an additional module — a platform for transparent and convenient work with big data, EW DataFlow. As a result of the implementation, we integrated the product with other parts of MMP and transferred the operator's big data processing to it.

"To create an MLOps system, operators usually use a set of open-source solutions that they customize for their company's infrastructure. Such development takes a minimum of six months and requires the presence of focused specialists for each tool in the team. The DataFlow platform helped us build the right MLOps system within a month. With the solution from Eastwind, we received an understandable tool integrated with all the systems we needed. Now, DataFlow allows data analysts to focus on their tasks without being distracted by the DevOps component."
 

Dmitry Kostenev, Head of Data Science Department, MegaFon Tajikistan

EW DataFlow capabilities

Eastwind's platform enables data scientists to perform all big data operations and calculations within the Hadoop ecosystem using a single interface.

— Integration with any analytical resources
— Python and Jupyter Notebook
— Real-time data and algorithm quality control
— Automatic translation of Python models into industrial code
— Setting up data processing pipelines from an external source to the end user
— Data Lineage visualization


On the main screen, DataFlow visualizes and monitors running data processing and analysis tasks

Eastwind integrates its data team with the customer's processes when implementing the platform. We help the client set up data processing flows, connect all sources, and implement AI/ML models to create a deep subscriber profile. Eastwind also provides consulting services for the rapid implementation of business cases based on big data and provides technical support 24/7/365.

 

Features of implementation in MegaFon Tajikistan

Migration of calculations to EW DataFlow

At the start of the implementation of DataFlow, the operator launched more than a hundred regular data processing tasks through the analytical module, the results of which were used for marketing campaigns. As a result of the migration, we transferred all calculations of analytical models to the new platform without stopping a single advertising campaign. MegaFon Tajikistan analysts were able to implement and improve ML models.
 

Increasing capacity by 50%

During the migration process, the customer expanded plans for using the new platform and asked to consider these changes in the already launched project. In the last month of implementation, the client, on our recommendation, increased the number of servers in the Hadoop cluster from four to six. This capacity will be sufficient to support calculations and to develop new operator operations.

"DataFlow is currently in active use at MegaFon Tajikistan. We maintain constant contact with the client through technical support and personal chats. We help resolve any issues and gain a deeper understanding capabilities of the product."
 

Vladimir Borkovsky, EW DataFlow Product Manager

Project development

MegaFon Tajikistan influences the development of the product. We regularly discuss the functionality of the installed platform and consider all customer requests in detail — some of them become part of the DataFlow roadmap. This is how we improve our solution based on real customer needs.

Learn more about DataFlow