Veriff Makes Identity Verification Simple Using Amazon SageMaker and Amazon Rekognition

Without a high-performing identity verification solution, businesses can face revenue loss, reputational damage, and noncompliance penalties. Veriff understands this importance and continually takes steps to improve its identity verification solution. As the company scaled its operations, it incorporated fully managed AI services from AWS to optimize the solution and improve the customer experience. This change improved efficiency, scalability, and innovation, helping Veriff foster trust in every online interaction.

“Up until recently, we used foundational AWS to power the infrastructure while running AI services internally,” says Rästas. “Now, we’re migrating to specific tools so that we can fully take advantage of what AWS has to offer.”

Veriff is using Amazon Rekognition to extract information and insights from images and videos, to complement its in-house facial recognition capabilities with high accuracy, subsecond latency, and global availability. To run deep learning models for document identity verification and authentication, the company adopted Amazon SageMaker, which enhances Veriff’s in-house technology to build, train, and deploy machine learning (ML) models for any use case.

One of Veriff’s most foundational ML models is responsible for detecting document specimens quickly and accurately. Before deploying this model, the process could take up to 2 weeks. Veriff reduced this time to less than a day by adopting multimodel endpoints for Amazon SageMaker, which provide a scalable and cost-effective solution for deploying large numbers of ML models.

Amazon SageMaker has become a key tool for cost-effective model training, and engineers can deploy multiple ML models on AWS, facilitating rapid experimentation. Using Amazon SageMaker for model training has reduced the cost by 27 percent compared with Veriff’s previous training environment.

“We have hundreds of different ML models that we simultaneously maintain and develop,” says Rästas. “Our team doesn’t need to worry about how many images or sessions go through the stack—with Amazon SageMaker, we can meet this scale quickly and cost effectively.”

Veriff is using Amazon Rekognition to compare faces from identity documents with selfies taken by the user to confirm that the individual matches the provided documentation. To bolster this process, the company has integrated various Face APIs from AWS.

Veriff can verify user faces within 1 second, exceeding its latency targets. The company’s customers and their users enjoy rapid and accurate identity verification. “Now we can scale our identity solution globally, minimizing bias and creating a seamless verification process for all,” says Geo Jolly, senior product manager of AI at Veriff.

On AWS, Veriff has achieved further enhancements in its identity verification workflows, resulting in heightened efficiency and performance. With cost-effective ML and AI services, it has saved resources that it can invest in new projects, accelerating its speed of innovation. Thanks to this evolution, users benefit from a more seamless and reliable verification process, an advantage that will continue to improve.

As Veriff looks to the future, it will refine its service to provide data scientists with richer data and facilitate the development of advanced solutions. Additionally, it is investigating the use of AWS services to identify and combat the increasing threat of deep-fake fraud, a technique that uses generative AI to produce convincing impersonations.

Veriff’s commitment to staying ahead of potential threats underscores its dedication to offering the most secure and cutting-edge solutions—a mission that it will continue to achieve on AWS.