Red Hat Summit
Turkcell, a digital telecommunications and technology service provider, sought to adapt to the growing importance of artificial intelligence (AI) for telecommunications by creating an AI service platform. Using Red Hat OpenShift and NVIDIA GPUs to create this infrastructure, Turkcell has established a responsive, cost-effective foundation for developing and delivering innovative AI-powered services faster. As a result, the company can remain an industry pioneer and take advantage of new revenue opportunities.
Benefits
- Created responsive infrastructure to deliver innovative AI applications faster
Cut provisioning times from months to seconds - Reduced AI development and operations costs by 70%
- Created inclusive data science platform
Simplifying development and deployment of AI workloads
Founded in 1994, Turkcell is a digital telecommunications operator headquartered in Turkey. The company offers a portfolio of digital services, as well as voice, messaging, data and IPTV services on its mobile and fixed networks, including fiber internet. Since launching the country’s first GSM mobile network, Turkcell has continuously innovated to provide new voice and data services to its customers through its parent and international subsidiary brands. It was one of the first global telecommunications operators that adopted LTE-Advanced and 3GPP carrier aggregation technology to enhance the bandwidth and rate of its services. Turkcell is the first operator in the region to offer live TV broadcasting over 5G and was the first to reach the fastest 5G connectivity rate using commercial terminals.
In 2018, Turkcell created strategic initiatives to embrace data science, the Internet of Things (IoT), blockchain, and artificial intelligence (AI). “With AI becoming fundamental to many digital applications, we decided to focus on creating AI-powered services to enhance our products and take advantage of new market opportunities,” said Osman Kiraz, Artificial Intelligence Industrial Solutions Manager, Turkcell.
To support this work with faster, more efficient development and deployment processes, Turkcell decided to replace its monolithic IT architecture with a microservices- and container-based environment.
Adopting a robust, scalable container platform
Turkcell decided to deploy Red Hat OpenShift for enterprise Kubernetes orchestration for several other projects at the company, as the foundation for its AI-powered application workloads. The platform runs in a private cloud environment on NVIDIA A100 Tensor Core GPUs. These GPUs help speed AI data analytics, modeling, and inferencing tasks. The Red Hat OpenShift NVIDIA GPU operator provides scheduling for these accelerated workloads.
“We’ve worked with containers for about four years. We wanted the benefits of a comprehensive Kubernetes platform—agility, resiliency, consistency, and portability—with the innovation and security of open source, coming from many people reviewing and enhancing the code,” said Ozan Orcunus, Senior Cloud Expert, Turkcell. “As one of the first companies to invest in the Kubernetes container ecosystem, Red Hat was the best vendor to help us develop, deploy, and run our AI workloads in containers.”
Turkcell’s team spent around eight months working closely with Red Hat Consulting and NVIDIA to plan, design, and implement the new architecture.
Building on familiar processes and resources to deliver AI innovation faster
Improved time to market for specialized AI workloads
With Red Hat OpenShift, Turkcell’s developers can use the same continuous integration and continuous delivery (CI/CD) processes used for Java™ applications to develop new AI applications. Self-service capabilities let developers and data scientists create new environments as needed. Any specialized infrastructure—such as drivers that connect applications with GPU processors—is provisioned and managed by the container platform, isolated by an abstraction layer to help Turkcell’s teams focus on innovation.
“Red Hat OpenShift allows us to easily deploy identical environments for development, testing,and production,” said Erkan Haspulat, Principal Artificial Intelligence Portfolio Architect, Turkcell.
Turkcell can now deliver its AI services to market faster, such as:
- Advanced Turkish-speaking speech-to-text chatbots.
- A text-to-speech translation service to help autistic children communicate.
- An image-to-speech service reads magazines out loud.
- The Life Box image processing service that recommends images for social media.
- The Fizy application that analyzes a selfie to suggest music based on detected mood.
One key project, the Spread the Smile campaign, uses facial-recognition technology to analyze selfies in Turkcell’s Digital Operator app to determine the size of food donations made to Yedikule animal shelter. This creative and innovative use of AI for social good earned Turkcell recognition as a finalist in the 2021 Shorty Awards for Artificial Intelligence, Creative Use of Technology.
Cut provisioning and scaling times from months to seconds
By adopting an automated, container-based infrastructure approach, Turkcell can provision its resources and environment dramatically faster than before. As a result, the company can not only deliver new services faster but also scale automatically to adapt to demand.
“A microservices approach with Kubernetes containers helps eliminate the challenges we saw with our monolithic architecture. We can divide our infrastructure into consumable resources to support new projects as needed,” said Orcunus. “And scaling doesn’t require any human intervention, because the platform already handles that work.”
