peshkova - stock.adobe.com

CTO interview: David Rafalovsky, Sberbank Group

Russia’s Sberbank is well into its digital transformation journey, with artificial intelligence and supercomputing now at the centre of its plans

David Rafalovsky is chief technology officer (CTO) at Sberbank, Russia’s largest financial and banking institution, which has now laid claim to being the country’s most innovative bank by reinventing itself as a tech company with a banking licence.

“Financial companies attempt to fulfil a specific demand for financial products and services, while tech companies try to fulfil very different needs of consumers, and we think of ourselves in that way,” Rafalovsky tells Computer Weekly.

“Trust me, we are not going to forget how to provide mortgages, consumer loans and other financial products to our 100 million customers, but we would like to offer additional products and services, making the consumer happier as a result.”

Rafalovsky stresses that, unlike traditional financial institutions, Sberbank has been investing a higher portion of its revenue in tech transformation and building more technological products.

“This requires more significant investment, including foundational technology investment which financial companies don’t normally do themselves – they rely on others to do it for them by leveraging off-the-shelf platforms,” he says.

Sberbank’s tech transformation has been going on for a while, and even the Covid-19 pandemic has not had a major impact on its plans.

While staff in the bank’s more than 14,000 physical branches still came to work during the lockdown to provide critical services for organisations, the vast majority of Sberbank employees had to switch to working remotely.

“Within weeks, we had to enable remote access for almost 100,000 employees who did not have it before – remote access not just for their email, but for their entire workspace, which, if they are a developer or a product designer, requires the use of quite a few tools to be productive,” says Rafalovsky.

“And we’ve actually seen an increase in productivity in some of our development teams working remotely.”

According to Rafalovsky, 70% of Sberbank’s employees who are currently working remotely will return to the office soon.

Read more about banking IT in Russia

The bank faced many challenges before Covid-19 took hold, including running its own digital platform for all services.

“Several years ago, Sberbank made an untypical decision to build its own digital platform for banking and non-banking services,” says Rafalovsky. “Very few companies in the world decide to build their own digital platform rather than buy one.

“Over the last two to three years, we have had quite a few challenges building this platform and proving that it works on our scale. Our scale – 350 centralised automated systems, more than 15,000 operations per second, and over 120PB of data – is quite an architectural and technological challenge, but we successfully overcame that.”

According to Rafalovsky, moving to agile methodology also presented significant challenges – both culturally, as employees were not used to working in this mode, and organisationally, given Sberbank’s scale.

“Currently, we have 2,500 concurrent agile teams working on various products,” he says. “It would have taken us years in the past to adjust our processes and our culture to operate in this environment.”

Another potential issue facing many tech companies that provide digital services is data security. “Cyber security is front and centre for every product design conversation, regardless of whether it is a banking or non-banking product,” says Rafalovsky.

“We had some issues with internal fraud last year, which we successfully overcame, but we’ve never been penetrated from outside, and that is a testament of how much effort, time and investment we dedicate to cyber security and protecting our customers’ data.”

Sberbank’s digital platform is based on a cloud-native architecture, says Rafalovsky. “We believe in this so much that we opened a standalone business, SberCloud, to become a prominent cloud provider in the Russian market. The cloud business in Russia is still immature and relatively small, but it is growing rapidly.”

AI promises

Sberbank considers artificial intelligence (AI) to be one of the most promising technologies. “Sberbank started to invest in AI long before it became sexy in the world and almost mandatory for being part of the conversation, investing not just in applied use cases, but in foundational AI research,” says Rafalovsky.

“AI is already embedded in every process and every product the bank offers to some degree. Of all of our technology stack, AI holds the highest degree of potential. We use AI because it’s useful, not because it’s trendy.”

Rafalovsky describes AI’s potential in customer services and several other use cases as “tremendous”, adding: “As much as we’ve achieved already, we are just at the beginning of this journey.”

One AI use case for Sberbank is a commercial financing product called K7M. This Russian abbreviation stands for “credit in seven minutes”, which means it takes just seven minutes between a customer applying for a loan and collecting the money. All banking processes, including anti-fraud and credit checking, are AI-based, which enables decisions to be made in minutes, rather than days using traditional processes.

Another area where Sberbank sees a lot of potential for AI is voice assistants. “Voice [interface] is all based on AI,” says Rafalovsky. “The ability of the voice assistant to recognise your demands and predict what you will need next is all based on AI.

“Voice interaction is a critical investment for us. In the next three to five years, we predict a significant uptick in the number of applications using voice, integrating with a voice interface or completely moving to a voice interface where it is feasible and organic.”

Christofari supercomputer

The use of AI has also pushed Sberbank into another tech area – supercomputers, says Rafalovsky. “Our usage pattern of AI capacity reached a certain level where we just needed more. We couldn’t find this capacity in the Russian market, so we partnered up with Nvidia to build a supercomputer for the purpose of AI model training.”

In November 2019, Sberbank unveiled its supercomputer, the Christofari, named after Nikolai Christofari, Sberbank’s first customer 178 years ago.

“We decided to make the Christofari a commercial offering,” says Rafalovsky. “We believe that parity and accessibility of AI capacity and development is a great accelerator of AI adoption in the Russian market.”

Read more on IT for financial services