Kytes Grows Content Platform Adoption by 40% with Generative AI on AWS

Established in 2021 Kytes is a platform that distributes a broad range of content types, including interactive reports, statements, videos, and eBooks, while harnessing artificial intelligence (AI) to enhance content. The company is a spinoff from the interactive digital textbook publishing platform Kitaboo, used by publishers globally.

Shortly after its launch, Kytes sought to incorporate generative AI to enhance customer experiences and extract value from its content. By using generative AI, the platform can summarize lengthy texts, swiftly extract key points, and provide AI chat assistant that responds within the context of the content and facilitate the creation of multilingual subtitles for video presentations. Additionally, it can bolster reader engagement and comprehension rates, tailoring questions to align with readers’ interests and the specific goals of customers. Nayan Goenka, head of product engineering at Kytes, says, “We see a huge opportunity to elevate the value of content and unlock possibilities using generative AI.”

To further develop its capabilities and stay competitive in an increasingly AI-driven world, Kytes engaged Amazon Web Services (AWS). Kytes was already using several AWS services, including Amazon SageMaker, to deliver and develop its applications. “We were aware of AWS launching generative AI services, and that sparked our interest," Goenka explains. "The advantage of AWS is that you don’t need all the answers to get started. We knew we would receive the support we needed."

Kytes participated in a generative AI session organized by AWS in Mumbai to explore the technology. Goenka and colleagues also engaged in workshops led by generative AI on AWS experts. This included delving into Amazon Bedrock, a fully managed service offering a range of high-performing foundation models for various generative AI capabilities. "We acquired sufficient knowledge to overcome the entry hurdle and start experimenting," recalls Goenka. 

The company deployed Amazon Bedrock on Amazon Elastic Compute Cloud (Amazon EC2) clusters, comprising instances with a mix of 2 and 4 cores and 4 and 8 gigabytes of memory. Kytes initially began its journey with Anthropic's Claude, a large language model (LLM) for generative AI applications, progressing through versions 1, 2, and 3 before finally adopting Claude 3 Sonnet.

As Kytes began developing its generative AI applications, it utilized indexing for natural language processing (NLP) to represent data in a machine-readable format. However, the AWS team suggested adopting the foundation model Amazon Titan, along with Amazon Titan Embeddings to convert natural language text, such as words or phrases, into numerical representations or vectors. "Once we grasped how vectors could produce faster, more accurate outputs, we transitioned to using Amazon Titan for indexing," explains Goenka. "That shift has enhanced our agility and efficiency in design."

Using Amazon Bedrock, Kytes effectively developed various generative AI applications for its e-learning platform. These applications include "Talk to My Book," text summarization, a multiple-choice question (MCQ) generator, and QuizMe, a learning assessor. With Talk to My Book, readers can engage in natural-language interactions with written content. The other applications have helped users summarize lengthy texts and assess comprehension levels through automated questionnaires and quizzes.

Kytes launched these applications around September 2023. With Amazon Bedrock and support from the AWS team, the company saved significant development time. "The time saved through Amazon Bedrock and our collaboration with AWS was crucial, allowing us to expedite our time to market. This advantage proved significant, especially considering the global surge in e-learning companies developing generative AI applications," says Goenka.

Following the launch of the applications, parent company Bookit Digital evaluated the performance of Amazon Bedrock against other generative AI-managed offerings it had tested. Subrat Mohanty, CEO of Bookit Digital, notes, “Compared to other generative AI solutions, Amazon Bedrock consumed 70 percent less DevOps resources and facilitated the development of features and completion of experiments 60 percent faster.”