A few years ago, Abhijit Bose was leading Facebook’s East Coast artificial intelligence research team, helping to teach computers to spot sunglasses on people’s faces. 

Now, he’s helping Capital One create A.I. systems that can more quickly spot ATM scams and recommend credit card rewards programs to customers.

Bose leads the company’s machine learning center, which develops A.I. tools for other data scientists at the bank. With these tools, Capital One can better understand which rewards programs would be more appealing to certain customers, for instance.

In an effort to improve its machine learning acumen, Capital One is on a hiring spree. The bank plans to add 3,000 additional technologists by the end of the year, with a focus on machine learning specialists and software engineers who can build and maintain the complex infrastructure for powering A.I. software.

Hiring that many techies will be a challenge, particularly machine learning experts, who are in demand all across the corporate world. If the company fails in its hiring goal, it may have to turn to contractors or delay some of its tech projects, Capital One chief information officer Rob Alexander said. 

Tech giants like Google parent Alphabet and Facebook offer top A.I. recruits who have Ph.D.s pay packages of up to $500,000 annually, said Alex Ren, CEO of A.I.-talent recruiting firm TalentSeer. Banks, insurance companies, and other financial services firms typically pay such recruits $300,000, making it difficult for those businesses to compete against the tech giants.

Additionally, because demand for A.I. talent is so high, machine learning engineers with just three years of experience can receive five different job offers within two months of their initial job search, Ren said. If uninterested in working at companies like Google or Facebook, job seekers may consider high-flying startups that can provide a huge windfall if and when they go public.

Story continues

Alexander acknowledged that hiring A.I. talent is difficult and declined to comment about whether Capital One will match the tech giants’ pay. But he said that his company tries to be competitive.

“Mostly we’re going head-to-head against other technology companies, more so than banks for the types of talent that we’re looking for,” Alexander said. “That’s the price of entry in the game, as you have to have a competitive value proposition—compensation is a big part of it. It’s not the only thing, but it’s definitely a big part of it.”

Capital One hopes to take advantage of a job market in which workers increasingly shop around, Alexander said.

Although Capital One intends to hire machine learning experts, it’s also looking for software developers who know how to use the cloud computing services of giants like Amazon Web Services, Bose said. In 2020, Capital One closed its last internal data center and moved all of its IT infrastructure to AWS.

Training a machine learning model to spot credit fraud can require using over 100 different graphics processing units, or GPUs, the equivalent of using 100 beefy personal computers to perform one task. The company also must create technology that helps machine learning training continue even if one of the cloud-based computers it’s using fails, a common problem.

Although Capital One could rely on the easier-to-use machine learning tools sold by the likes of Amazon, Bose and Alexander said those tools lack the power of their company’s in-house equivalents. Customers also increasingly want speedy financial apps, and any lag time can cause them to seek alternatives. 

“You can imagine our whole business is about making decisions, and every one of these decisions is better if you can make the decision in real time with more data and better algorithms,” Alexander said.

While it may seem dramatic, Alexander said Capital One must invest heavily in machine learning and related technology. Just like Uber and Airbnb disrupted incumbents in the transportation and hospitality sectors, financial services firms face a risk.

Said Alexander, “The digital revolution is an existential threat to banks in a new way, and if we are going to be a winner in this business, we must look like and operate like a great technology company.”

More must-read tech coverage from Fortune:

  • A Tesla “recall” applies to nearly every car the company has sold in China

  • Post-Brexit data transfers from the EU to the U.K. are safe—for now

  • Semiconductors are in short supply. 3 experts weigh in on what to do about it

  • Southeast Asia’s most valuable company has an edge in the battle of the super-apps: Its gaming business

  • 11 smart, time-efficient podcast tips and tricks

This story was originally featured on Fortune.com

(305) 707 0888