New version carries far more automation power
Version 1 of Paradatec’s automated document processing technology came out in 1997, nearly 25 years ago, when the mortgage and real estate technology operation was still based in Germany.
The technology was ahead of its time even then, argues Mark Tinkham (pictured), the company’s director of business alliances, starting with an optical character recognition engine invented nearly 10 years before that.
“That was the foundation of the original product. We are uniquely able to take advantage of the ability to capture every word and every character on every page in less than a second per core,” Tinkham said. “It scales very highly and very fast. Nobody else has been able to touch us since 1997 with that approach.”
Paradatic is a provider of AI-based document classification and data extraction technology for mortgage loan processing. Earlier in April, it released version 8.0 of that platform – known as AI-Cloud, which is hosted by Amazon Web Services. New features include automated document processing technology and expanded data audit trails.
The company said at the time of its early April announcement that the revamped platform is the result of a multi-year effort to achieve completely automated document classification technology. The core platform has historically relied, in part, on a textual analysis tool based on artificial intelligence (AI), which is trained repeatedly against thousands of copies of different mortgage documents. It is updated regularly.
“Textual analysis is a beautiful thing that allows us to take an approach that nobody else can,” Tinkham said.
The technology governing textual analysis depended initially on neural networks – kind of an early version of AI.
Visual classification tech
For the new version of AI-Cloud visual classification capability has been added to the product, involving some of the latest AI technology, Tinkham said.
“We’ve hired on some folks from some leading educational institutions to help us develop this new addition to our product,” he said.
Alongside visual classification capability, document boundary technology is also now part of the mix, with an ability to determine page boundaries.
“There is a challenge with document indexing in not only determining what the documents are but determining when they break, and you get a whole bunch of contiguous order,” Tinkham said.
Textual analysis, classification capability and document boundary technology have helped create a better tool, which, he said, “has really rocketed our automation and accuracy numbers.”
The platform has been in the cloud since late 2021, something Tinkham said has broadened the company’s ability to reach smaller lenders.
“It opens up the ability to walk away from all the hardware management and overhead that’s involved with setting up our technology” on the premises, Tinkham said.
Beyond that, however, being on the cloud has added flexibility for both client and customer alike.
“It allows us to provide our technology and make it available with our APIs to many more lenders and servicers and industry players,” he said. “A lot of times, having the resources and capabilities to set up your own environment and so forth, it’s just a deal stopper. That’s no longer the case with AI Cloud.”
Paradatec was founded in Germany in 1988, long before anyone was talking about neural networks in AI. Now, the company still has research and development operations there, though it is now primarily an independent subsidiary based in Cincinnati.
About 45 employees work for the company at this point, domestically and in Germany. Beyond the company’s US mortgage industry business, it also has clients in Europe and the UK, but the focus overseas is mostly on the insurance space.
In the US, Paradatec’s customers include three of the 10 largest US banks and four of the 10 largest mortgage servicers, according to the company.
CoreLogic, Northpointe Bank, and Guaranty Home Mortgage are among the company’s domestic customers.
Paradatec has been around for a long time, and Tinkham acknowledges the reality of competition, but he said it hasn’t made much of a negative impact on the company’s business.
“We haven’t seen any real competition that’s hurt us at any way at this point,” Tinkham said. “The marketing out there and the AI this and the AI that, is sometimes a little amusing. But we’re still gaining sales and we run into all this competition, and they fall away very quickly.”
Tinkham said part of the reason Paradatec has done well is that the mortgage space is difficult top operate in.
“You see a lot of folks come and go in the mortgage space. It’s a tough space to do what we do,” Tinkham said. “You need a special technology to do it.”