To be a leader in telecom, you need to lead in AI: Chandru Sargor

He tells that Artificial Intelligence (AI) is solving real life challenges in telecom networks

From cell decongestion, to coverage, Chandru Sargor, head of Ericsson India's Global Artificial Intelligence Accelerator (GAIA) and Ericsson R&D Site, Bangalore, says that Artificial Intelligence (AI) is solving real life challenges in telecom networks. In an interaction with Subhayan Chakraborty, he also explains how Agentic AI will be the way forward. Edited excerpts: 

   

Which are the areas within telecom with the highest incidence of AI adoption?

 

 

In every business unit within Ericsson, whether it’s on the Mobile Networks side, where we’re talking about RAN, Transport or Core Networks, or on the Enterprise Networks side, wireless LAN and private cellular networks, across the board, we are seeing adoption of AI. It is truly a technology horizontal that cuts across all the areas. 

 

The understanding is very clear that to be a leader in telecom, you must also lead with AI capabilities. We are working to translate advanced AI technologies into impactful solutions for telecom networks. 

 

Why is GAIA important for Ericsson and India? 

GAIA was started over five years ago to accelerate the adoption of AI and ML across Ericsson’s portfolio. It was established as a central organisation within Ericsson to be best positioned to support all the business units, and to see how we can bring AI into their products and services. This has given us a lot of perspective over the years about what the business units are trying to accomplish using AI, what common themes are emerging across their AI objectives, and how we can harmonise these efforts to create common, reusable AI assets that can be applied across multiple areas.

 

We are primarily located in three geographies: North America (in the US, Santa Clara, and Canada, Montreal), Sweden, in Stockholm (which is our headquarters), and in India, we are based in Bengaluru and Chennai. Overall, about 170 data scientists and data engineers. About half of them are based out of India.

 

What are some recent innovations at GAIA?  GAIA had initially worked in the traditional AI space, integrating traditional AI/ML models into Ericsson’s portfolio. Over the last couple of years, large language models and Generative AI (GenAI) have really come to the forefront. Ericsson has invested in language models that are fine tuned for telecom. We have built a GenAI platform, optimised for telecom, that we call the Ericsson Language Intelligence (ELI). This platform will be leveraged to deliver innovative Gen AI use cases built on top of it. A demo at the India Mobile Congress, called the Generative AI-driven Network Autonomy, leveraged our GenAI platform, coupled with advancements in Agentic AI technology, to transform future network management and operations. 

 

What about research in telecom networks in particular?  Managing and operating telecom networks is challenging because they have become extremely complex. GenAI and Agentic AI really have the potential to significantly transform this. Agentic AI technology gives you the capability to deploy several AI agents throughout the network, each of them optimised for a specific task. These AI agents communicate with one another, and given a problem in the network, can collaborate to find the right solution. They can do this because generative AI-based agents have the ability to look at information such as telecom standards documentation, product documentation, trouble reports, etc., and build up their own knowledge base from that. 

   

How is AI helping beyond the core technological aspects? 

AI has been used to solve some real-life problems. One of the challenges that we often encounter in cellular is coverage. The problem is, when you have antennas deployed on towers, there’s something called an electrical tilt, to enhance cellular coverage and minimise interference. In the past, any changes to the tilt configuration would typically be infrequent because it had to be done manually. With AI, the algorithm will decide what the optimal electrical tilt should be automatically, based on current network conditions, and so, we're able to dynamically optimise coverage and power, thereby benefiting end users. 

Another example - sometimes, a certain cell may become congested, therefore, the users on that cell will experience a degradation of their services. We used AI to predict, ahead of time, when a given cell was likely to become congested and then 

proactively move users from that cell to a different, less-loaded cell. This way, we were able to optimise the overall network performance across cells. This improved the parameters such as users’ throughput and 

reduced call drops.