These were the realizations we came to in 2021 after attempting to deploy a generic AI and data platform within Rakuten Mobile. Adoption was nonexistent because the telecom experts tasked with running network operations around the clock were too busy to learn data science.
In reality, this was not a surprise. Telecom experts were too busy planning network rollouts to automate what they were doing in 2018. In 2019 they were too busy building and in 2020, too busy operating.
They will always be too busy.
That is when we realized we had to make the tooling integrate to the experts and not the other way around.
We did this successfully with automation and now, with AI.
The results of everything we’ve learned, developed and fine-tuned are represented in the Rakuten AI for Telecom platform.
Available to all operators globally, it leverages two years of deployment learnings to fast-track AI-powered improvements specifically within telecom domains. It brings experts closer than ever to being able to maximize the effectiveness of every minute spent operating the network.
The harsh realities of complexity, competition and cost continue to plague telecom. Hypothetically, automation and AI can introduce the speed and efficiency required to overcome these challenges.
While AI is set to power the next level of automation, the biggest barrier to success is a lack of AI experts. There is a global shortage of AI experts and the ones that do exist are not lining up to join telecom.
That’s why Rakuten AI for Telecom was not built to enable data scientists to academically make networks smarter. Rather, it was designed to help the existing telecom experts of today close skill gaps associated with supporting the realities of managing complex, automated mobile networks at scale.
Our goal with Rakuten AI for Telecom is to completely decouple telecom experts from needing to be data scientists before they could effectively interact with AI models.
Case in point: For the past two years, 10 data scientists have supported more than 300 telecom experts who have interacted daily with Rakuten AI for Telecom, creating new models, moving them safely into production and continuously refining performance with more parameters, more data, and more learning.
A list of example use cases and models can be found here for everything from Remote Electrical Tilt to Sleeping Cells. Only ten data scientists are required to support these hundreds of experts supporting the network at full scale.
Our approach has strong parallels to what Hugging Face has achieved in simplifying accessibility and deployment of AI models. Just as Hugging Face helps users easily integrate advanced machine learning capabilities into applications, Rakuten is empowering telecom experts to seamlessly implement, collaborate around and scale sophisticated AI models.
Key features on this front include:
In addition to being easy to use, Rakuten AI for Telecom is also extremely light touch. The initial platform can be deployed by a handful of engineers in merely hours.
Now, teams can publish model requests, deploy use cases and achieve business results faster. The impact is more quickly observed by exec teams.
It is our intention to democratize advanced AI-powered telecom capabilities and make them accessible to a wide range of users. In turn, we can empower mobile operators to completely control how they deploy AI, tailoring models for their specific requirements or problems.
This control extends to how the data that powers AI is managed and stored.
Rakuten AI for Telecom’s architecture supports federated data governance and distributed data management for maximum efficiency and scalability. This helps reduce AI model deployment times while ensuring governance and security.
Operators also maintain complete control over their data and AI processes with robust governance and security capabilities.
Rakuten AI for Telecom is more than a platform. It is a tool mobile operators can immediately deploy to bridge AI skills gaps and set the direction of an AI future they control.
