Achieving Massive Business Growth with Large-Scale Models in Production

Wargaming is an award-winning online game developer and publisher, and one of the leaders in the free-to-play MMO market. The company delivers authentic gaming experiences and services across PC, console, and mobile platforms. With over 110 million players worldwide, 15+ game titles, and almost 2PB of data, Wargaming has successfully scaled AI that has driven exponential growth over the last few years. Their data science and engineering teams have 1,500+ models running in production today that power dynamic in-game experiences, improve support efficiency, and deliver advanced reporting systems that drive critical business decisions.

Wargaming, an award-winning online game developer and publisher, faced a significant challenge in scaling their AI across business units. With over 110 million players worldwide, 15+ game titles, and almost 2PB of data, the company had 1,500+ models running in production on a single server solution. This severely limited infrastructure and constrained data scientists, as the overhead cost of adding servers was extremely high due to per core licensing. The existing platform also limited data scientists to a language and packages pre-approved by the platform. Wargaming needed a solution that could support large-scale models in production, scale their servers, minimize overhead costs, and provide flexibility for their data scientists.

Wargaming adopted cnvrg.io, a modern MLOps automation platform, to support their massive amount of models in production. This platform provided fully automated pipelines across their product, support, and business reporting units. Compared to their legacy platform, cnvrg.io delivered a modern Kubernetes and cloud native based infrastructure that easily blended with existing IT and offered the ability to grow without limitations. Wargaming was able to utilize the powerful machines they already had, and easily add acceleration and powerful servers to achieve even better performance. The data scientists were able to quickly adapt to cnvrg.io with an intuitive UI that gave them the flexibility to use their preferred language, add customized packages and use other open source capabilities.

• By adopting cnvrg.io, Wargaming gained flexibility and scale that they never could with their legacy platform. The platform enabled Wargaming to fully automate 1500 models in production while accelerating continuous application delivery. They achieved results at the same performance in half the time and at half the cost of their legacy platform. cnvrg.io enabled them to seamlessly adopt new servers and accelerators to improve performance. They saved money and time by not having to purchase and install a new bare metal machine from their previous vendor. Now, they can scale seamlessly to a hybrid cloud environment, add accelerators and grow to maximize performance. They have accelerated their ML pipeline by more than 50% by eliminating delays from ML platform developers. Now, data scientists can run their entire ML workflow autonomously, without DevOps or platform developers. In addition, Wargaming has achieved a nearly impossible goal, approval from the data science team. For the first time, Wargaming’s critical team of data scientists are lining up to use cnvrg.io.

• Reduced cost by more than 50% on servers and licensing

• Accelerated research to production from 3 months to 2 months

• Accelerated time to production by 30% by eliminating bottlenecks