Urbanbase launches services 20x faster with AWS

In its early days, Urbanbase built and operated a service platform in the cloud with the help of Amazon Web Services’ AWS Startups support program. This enabled it to automatically convert 2D drawings into 3D in seconds, render and decorate virtual interiors with 3D data, and create augmented reality (AR) experiences based on spatial and photo information.

The global demand for 3D/VR spatial related services is burgeoning, and to support its overseas expansion goals, Urbanbase needed the large-capacity web services that microservice architecture could best provide.

"We used monolithic architecture for our service platform, which was not an issue when it was small. But as the company grew, we needed to construct and operate larger systems, so we required a new architecture," says Bang HyunWoo, Urbanbase CTO.

Changing the architecture of a proven service platform isn’t easy, as migration problems may affect both customers and partners. So Urbanbase had a simple strategy: the existing system would upgrade to container-based environment, and the new service would be based on serverless architecture.

Urbanbase advanced its existing VM environment using Amazon Elastic Compute Cloud (Amazon EC2), a high-performance container orchestration service that uses computer vision and deep learning technology to provide spatial analysis, object recognition, user taste analysis, and product recommendations.

“For startups like us, it’s important to develop fast prototypes. Microservice architecture helps us realize new ideas steadily," Mr Bang says. "We didn’t need to waste time designing the model because we used the algorithm included in Amazon SageMaker. Modelling, training, and endpoint creation were also very simple, so we could develop a fully operational serverless front-end application."

Switching to serverless architecture enabled Urbanbase to develop and operate new service platforms faster and more efficiently. By combining Amazon Simple Storage Service (Amazon S3), AWS Lambda, and Amazon SageMaker, the development for spatial analysis platform was fast, efficient, and smooth. Developers no longer need to worry about infrastructure, and this allowed Urbanbase to provide a consistent development and operating environment.

“For development that requires AI functions, performance is very important – as model training time increases, development time increases. Compared to on-premise workstations, Amazon SageMaker’s performance is about 100 times faster,” Mr Bang explains. “Its flexible configuration also enables us to easily develop and train models and use CPUs selectively for inferencing, which helps reduce costs.”

“If we’d built infrastructures in a traditional way, it would have taken more than 20 times longer than using Amazon SageMaker in a serverless environment,” Mr Bang adds.