Iternal Technologies Improves Marketing ROI 30x with Generative AI on AWS

Iternal was founded to revolutionize marketing with an idea pulled from the history books: the assembly line. Rather than investing time and energy into creating assets that will no longer be useful after 6 months, Iternal set out to create personalized, relevant assets using modular blocks of content that are curated and combined by generative AI. To achieve this futuristic vision, Iternal needs reliable access to computing power at scale. The company chose to build its product suite directly on AWS to access the compute resources it requires. “The value of using AWS was to accelerate development on innovative technologies and bring them to a global set of enterprise customers who need multiregional availability,” says Williams.

Because Iternal is part of Inception, it was among the first to gain access to NVIDIA Riva voice cloning capabilities. It then used Amazon Elastic Compute Cloud (Amazon EC2), a broad and deep compute service, to get a proof-of-concept generative AI voice product up and running in 2 weeks. More specifically, Iternal used Amazon EC2 P3 Instances, which have up to 8 NVIDIA V100 Tensor Core GPUs for machine learning and high-performance computing applications, to power the solution. “We move very fast on AWS,” says Hanby. “We deploy at least four new enterprise-grade products a year.”

Iternal wanted to build on AWS because it needs to scale at will when training highly complex models to deliver accurate marketing assets. Its voice cloning initiative was a case in point. As the voice model training grew more compute intensive, Iternal had to spin down its old cluster, which used Amazon EC2 P3.2xl instances, and deploy a new set of infrastructure that had four times the memory using Amazon EC2 P3.16xl instances. Within 90 minutes, the company had spun up the new system and completed model training. “The ability to iterate and deliver capabilities faster when we’re not exactly sure how much compute is required is very valuable,” says Hanby. As of 2023, Iternal uses those capabilities to offer a near-real-time AI voice product with a latency of merely 300 milliseconds.

Iternal provides automation and generative AI technology in more than 300 unique language locales around the world, so the company must meet a wide variety of infrastructure requirements across its deliverables. It uses AWS services to meet those requirements. For instance, Iternal chose Amazon Elastic Kubernetes Service (Amazon EKS), a managed Kubernetes service, to run Kubernetes in the cloud. “The combination of Amazon EKS backed by Amazon EC2 P3 Instances gives our team a full-service suite of tools so that it can focus on building innovative products,” says Williams.

Speed and ease of use are critical for Iternal and its customers. The company delivers products that are ready to use out of the box by sales and marketing professionals who prefer to focus on deals rather than technology. The goal is to provide immediate value to use cases, regardless of who the user is. “We effectively empower enterprise organizations to scale by having 1 person do the work of 30 people at a tenth of the legacy cost,” says Hanby. That means a 30-times improvement in return on investment (ROI) as well as a 90 percent cost savings.

Iternal wants its solution not only to be straightforward to use but also to deploy quickly at the enterprise level. “Our average customer time to value is less than 14 days,” says Hanby. By harnessing AWS computing power, Iternal can deploy its AI offering to a new customer in as little as 30 minutes. Iternal then takes the time it saves on development and deployment and reallocates it to innovation.