Stride Learning Cultivates Student Creativity Using Stability AI in Amazon Bedrock

As an early pioneer in online learning, Stride has evolved over 25 years to become a leading lifelong learning company offering K–12 education, career learning, professional skills training, and talent development. Public school systems have begun leaning into the use of AI, indicating public interest in the new technology. In August 2023, Stride kicked off an internal hackathon, challenging its teams to create new product ideas using generative AI to increase solution development and empower learners. One of these ideas was built into Legend Library, an online game that makes it possible for learners to create their own imagery for stories using state-of-the-art generative AI technology. Legend Library was chosen because it could personalize learning and create more engaging experiences for students using text-to-image models, which take natural language descriptive prompts and turn them into images.

When building Legend Library, Stride needed to make sure its solutions considered state and federal education policies while still meeting the needs of its customers. To meet security and regulation requirements, Stride created an internal acceptable use policy for AI tools, vetted by its InfoSec and legal teams, to make sure it is not gathering, storing, or sharing sensitive student information. “We want to make sure students know how to interact with technology like generative AI responsibly and effectively,” says Will Plump, Senior Manager of Learning Innovations at Stride. “It’s valuable to make sure students have a safe way to engage with the most innovative technologies out there.” Using Amazon Bedrock, Stride has full control over the data used to customize the foundation models for Legend Library.

In addition to these challenges, the previous tool Stride used for image generation had issues with prompt adherence—such as nonsensical text and deformed images—which had the potential to cause confusion for students. When the company was looking for a generative AI solution, solutions architects at AWS brought their attention to Amazon Bedrock to streamline development for engineering teams. In order to meet the project requirements of throughput and prompt adherence, Stride chose Stability AI’s Stable Diffusion models in Amazon Bedrock as the best solution.

Working alongside AWS and Stability AI, Stride created a service endpoint for the early stages of development in the Legend Library project. “Amazon Bedrock and Stable Diffusion are the cornerstone of our application,” says Ernie Grey, Director of Platform Engineering at Stride. “The beta model of Stable Diffusion offered better results than our previous solution, and we were able to experiment with that rapidly using AWS.” Stride launched a minimum viable product version of Legend Library in February 2024, 6 months after the idea won the internal hackathon. K–12 educators and students can access Legend Library through Stride’s learning hub.

The Legend Library game takes prompts filled in by students and then uses generative AI to turn those into stories, images, and personalized storyboards. “We want to make reading personalized, engaging, and renewable for students,” says Plump. At the end of every story, students then answer several reading comprehension questions based on their grade level—in this case, K–4. Stride’s teachers were particularly fond of the comprehension questions as they sparked motivation and engagement with reading.

After Stride upgraded to Stable Diffusion in Amazon Bedrock, Legend Library produced higher-quality images that were more consistent and did not have warped or misshapen images. Using Stable Diffusion in Amazon Bedrock also meets Stride’s needs for data privacy, which is vital when dealing with student information. The combination of Stability AI and Bedrock maintains privacy using model level and platform level mitigations such as Amazon Bedrock Guardrails, used to implement safeguards customized to application requirements and responsible AI policies.

By using Amazon Bedrock, Stride has access to a variety of models all in one place, making development and speed to market more convenient for the company. “Our developers using AWS can spin up components in a sandbox environment, build prototypes, and use convenient tools on an AWS console to experiment,” says Grey. By using Amazon Bedrock and other AWS serverless solutions, Stride was able to achieve a throughput of 1,000 images per minute.

Next on Stride’s road map is to launch its Legend Library for a wider audience. It plans to develop an app available directly to customers and learners. Stride also plans to expand Legend Library to more grade levels, including the middle school grades. The company intends to continue innovating for the project using AWS, expanding its use and improving functionality for students.

“Within the educational landscape, it is important that students are using products in a safe way,” says Grey. “AI has a tremendous potential to delight students, generate curriculum, and provide instruction. It is going to be a bright frontier.”