AWS Telecom GM sees less need for central data lakes with gen AI

Chivas Nambiar is the general manager of the global telecom business unit at Amazon Web Services (AWS). Prior to joining AWS in 2020, he was with Verizon, involved in the carrier’s cloud and platform engineering work.

Nambiar spoke to reporters today about some of the things AWS will be talking about at the MWC trade show in Barcelona next week. And surprise! The company plans to talk a lot about artificial intelligence (AI), especially generative AI, which relies on constant input from users to keep optimizing the software.

Just yesterday AWS announced that BT Group is advancing its product development with AWS’ CodeWhisperer, which uses AI to assist developers in their coding.

According to BT and AWS, more than 100,000 lines of code have already been generated by CodeWhisperer in the first four months since BT has been using the product.

“They’ve already seen 12% of the code that developers are creating being auto generated,” said Nambiar. “They’re excited to see what rolling that out across 12,000 developers will do as they get more comfortable with the tooling.”

AWS says that CodeWhisperer automates a lot of the “tedious, repetitive, and time-consuming work done by software engineers.” One thing the tool does is make code suggestions, which sounds kind of similar to the suggestions we all see on our word-processing programs or our SMS systems these days. The BT programmers who are using CodeWhisperer get about 15-20 suggestions of code per day – with an acceptance rate of 37%.

Cox Communications chatbot

Nambiar said that the telecom industry has been collecting a ton of data for decades. “Having been in this industry for a while, I’m excited that we’re finally able to go back and look at all the data that we’ve generated.”

The data encompasses everything from telemetry, personal usage and customer journey. “All of these different kinds of data are living in many different areas,” he said. “We’ve struggled in kind of the classic AI techniques to understand where this data lives and bring it together in a way that’s clean and usable. And we’ve had this ambition to use it, but it’s been so hard to use it in the past. With early adopters now we’re starting to see that’s changing.”

For instance, it helped Cox Communications build a Q&A chatbot for field service technicians, who spend their days maintaining and repairing critical network infrastructure. 

The chatbot aggregates across multiple operational systems of Cox, and it’s built on top of AWS. Nambiar said that previously, Cox had automation in many different places. But the aggregation of that data sometimes took multiple hours. They built a solution on top of AWS Bedrock and Sagemaker where they could — across the sources of data — help technicians ask the right questions of that data, and then it provides a consolidated set of understanding of what is happening in the environment.

“We saw Cox being able to reduce the time from multiple hours to minutes in some of these complex use cases,” said Nambiar.

Data lake demand may drop

Recently, Rahul Atri with Rakuten Symphony said the company was fortuitous a few years ago when it established Rakuten Mobile, and later, Rakuten Symphony, because it made the decision from the get-go to keep all data in a massive, common data lake. And that has made gen AI model training so much easier.

But most traditional telcos have not had that luxury, and their data is stored in many different siloes – like the example given of Cox Communications.

Nambiar said he thinks there’s been a shift in the concept of data lakes over time. “There was, about three to five years ago, this model that all data needed to live in one data lake. Operators spent a lot of time trying to get this data centralized in one location. Over the last couple of years though, that model has really started to shift, not just in the telco world but across industries.”

He said there’s a concept of “data mesh” where data gets produced in certain areas but it can be accessed in a safe and secure manner. Then gen AI can start to query that data. And it’s in a much less structured way than before. “If I fast forward two to three years, I suspect we will have less focus on everything needing to come into one place."

He said the Cox Communications use case is an example of data in many different places, which is able to be queried anyway.

Follow our coverage of Mobile World Congress 2024 in our dedicated channel here.