TensorGo cuts costs by 40% and achieves real-time AI prediction models with OCI

TensorGo is an enterprise-grade, low-code deep-learning and computer vision platform, designed to help businesses build and deploy complex machine learning applications in a hassle-free manner. The business was created to solve some of the most challenging problems in the world, such as fraud detection.

TensorGo’s platform uses advanced computer vision technologies to solve critical business problems across a variety of industries. In one case, it uses computer vision to understand emotions, gestures, facial expressions, eyeball movements, as well as language use and tone to predict an individual’s emotional or physical state. This has many real-world applications, including assessing audience engagement; improving the diagnosis of dementia, depression, and other mental health issues; understanding consumer intentions and emotions; or even detecting signs of distracted driving.

With the easy extensibility of its low-code platform and the variety of use cases for computer vision to solve challenging business problems, TensorGo’s existing infrastructure could not meet the scalability and price-performance requirements to support its rapidly expanding business.

Positioned at the cutting edge of computer vision technology, TensorGo sought infrastructure support that would include superior price-performance and scalability for highly complex AI training and inferencing.

In its search for new infrastructure, TensorGo identified four key criteria to help weigh the options: superior price-performance and scalability for AI training and inferencing, exceptional technical support and true partnership from the vendor, improved performance at massive storage scale, advanced computing power for rapid data analysis, and lastly, a well-established history of supporting customers ranging from startups to global enterprises.

After a careful evaluation of the leading infrastructure solutions in the market, TensorGo found that only Oracle Cloud Infrastructure (OCI) would meet its complex needs. As a test, it used NVIDIA GPUs running on an Oracle Container Engine for Kubernetes cluster to deploy its machine learning and deep learning models. TensorGo achieved real-time or near real-time AI inference, as well as highly available data analytics to support its platform with the large data sets it needed for advanced AI training models. This persuaded the company that Oracle would provide the platform it needed to support the ongoing demands of its customers with cutting-edge computer vision applications to solve a variety of business challenges.

After TensorGo migrated from another cloud provider to Oracle Cloud Infrastructure (OCI), the company began using OCI Compute HPC GPU instances, which reduced its IaaS and PaaS costs by 40%. It also gained real-time and near real-time AI inference as well as highly available data analytics. The combination of NVIDIA GPUs running on OCI helped the company reach its goal of an infrastructure with exceptional price-performance and scalability with a partner who could support its needs, and those of its customers, in a rapidly changing industry.

Outside of the infrastructure support, TensorGo found other benefits from working with Oracle. As the operator of a platform as a service for many businesses, TensorGo must manage API access and subscription details for all users. The company adopted Oracle MySQL Database Service on OCI to store subscription details and API key pairs for customers, helping to ensure customers would continue to receive the service they expected.