AI-powered deep neural networks increase accuracy of credit score predictions


All Transform 2021 sessions are available on demand now. Look now.

Credit Karma has over 110 million users and a 90% customer approval rating, but it hasn’t always been that way. When the company launched 14 years ago, its approval percentages were single digits, CTO Ryan Graciano said at VentureBeat’s Transform 2021 virtual conference last week.

The reason for this turnaround? Big data and machine learning.

When Credit Karma launched in 2007, the company relied on traditional data centers because the cloud was not yet part of the conversation. There would have been issues with banking partners and credit bureaus, and “the compliance people wouldn’t even let you in,” Graciano said.

The company became very proficient at purchasing hardware and managing systems, but realized that physical hardware was limited.

“The problem with big data and the cloud is that big data moves very quickly, [and] technologies change very quickly, ”said Graciano. “If you need to go through a six to nine month supply cycle of materials [and] a major platform change, you’re going to be pretty far behind the curve.

This was the first problem Credit Karma sought to solve – the business needed more elasticity. It wasn’t just the time required to set up the hardware, but the fact that the hardware requirements were changing quickly to keep up with the new capabilities and the tech stack couldn’t keep up.

Credit Karma chose Google Cloud and its machine learning offerings because BigQuery and TensorFlow made big data management easier.

The evolution of machine learning

Attempts at machine learning were initially very straightforward. The company applied simple linear regression models to anonymized data in its databases. Later, Credit Karma switched to using gradient trees. Today, the company relies on large and deep neural networks to predict which banks will approve customers and at what rates. This technique performs about 80% of Credit Karma’s methods and helps facilitate Darwin, an internal system of experimentation and problem-solving.

The platform built by Credit Karma is reusable, said Graciano. There was a recommendation engine on top of the machine learning platform, and everything else was connected to it. Everything that happened with Credit Karma was coming from the system, whether it was receiving an email from Credit Karma, a push notification, or badges on the site.

“All of these things are powered by this one system. And so that allowed us to spend a lot of time on the nuts and bolts of how our data scientists work in the system, ”said Graciano.

It is much easier to add new data sources and clean the data than to define new algorithms. One way to improve the system is to add orthogonal data, rather than innovating on the algorithm, said Graciano. The company’s prediction capabilities expanded as new data sources were added.

“Getting these extra pieces of stuff is actually a lot more powerful than the 32nd iteration of our algorithm could ever be,” said Graciano.

Graciano admitted that it took a while to figure out what Credit Karma needed, such as a platform for data scientists to automate recycling models.

“I would say we’ve stumbled upon a lot of issues,” said Graciano.

Cloud was the way to go

Graciano recommends that companies move towards the cloud, as it increases interoperability within the external ecosystem.

“If you are looking for enhancement, you will usually get more enhancement by adding orthogonal data than by innovating on your algorithm,” he said. For Credit Karma, this was a strategic decision that paid off for the longevity of the platform, enabling it to amass valuable data and enabling the company to leverage it.

“Nothing is more strategic for us than data and having a lot of power over our data,” said Graciano. Many companies are likely to take this step for the same reasons, moving from a deterministic way of developing software to a more experimental setting.


VentureBeat’s mission is to be a digital public place for technical decision-makers to learn about transformative technology and conduct transactions. Our site provides essential information on data technologies and strategies to guide you in managing your organizations. We invite you to become a member of our community, to access:

  • up-to-date information on the topics that interest you
  • our newsletters
  • Closed thought leader content and discounted access to our popular events, such as Transform 2021: Learn more
  • networking features, and more

Become a member


About Author

Comments are closed.