Case Study | Amazon Bedrock GenAI Solution Bridges Structured and Unstructured School Data | Peachjar

About Peachjar

Peachjar is a technology company that provides a digital flyer management and distribution platform for K–12 school districts. The platform replaces traditional paper flyers with a centralized digital system that allows schools to communicate events, activities, and community programs to parents and guardians. Peachjar also captures engagement metrics that help schools and districts understand how their communications are performing. As Peachjar’s customer base and usage have grown, so has the volume of flyer content and associated data available for analysis.

Website: ms.peachjar.com

Challenges

Peachjar needed a solution that could quickly validate GenAI use cases, assess the capabilities of its existing data, and deliver results without spending months on planning, experimentation, or building custom systems from scratch. The goal was to understand what value could be unlocked from the data they already had.

Flyer content and flyer metrics were stored separately, with flyers residing in object storage and metrics maintained in a structured analytics environment. This separation made it difficult to extract insights that spanned both content and performance. As a result, it was harder to surface information about available events to parents or to provide school marketing teams with accessible engagement metrics.

As Peachjar continues to grow and serve more districts, improving access to this information becomes increasingly important. Without addressing these challenges, valuable flyer data and metrics would remain underutilized, limiting Peachjar’s ability to enhance community engagement and slowing progress on its broader AI journey.
 

Why AWS & Stratus10

AWS was selected because it provides a comprehensive set of services that align well with Peachjar’s existing data architecture. Amazon Bedrock offered a managed way to experiment with GenAI capabilities, including Knowledge Bases, without requiring Peachjar to manage or host foundation models directly.

In addition, AWS services such as Amazon S3, Amazon Redshift, IAM, and Secrets Manager made it possible to integrate GenAI capabilities into the existing environment while maintaining security, scalability, and operational consistency.

Stratus10 was uniquely qualified to help Peachjar evaluate and implement GenAI in a practical, production-oriented way. The team brought experience with Amazon Bedrock, AWS data services, and Infrastructure as Code, allowing the solution to be designed and deployed efficiently.

Peachjar engaged Stratus10 to move quickly from concept to implementation and to focus on delivering a working solution that could demonstrate real value. Stratus10’s ability to design a secure, reproducible solution aligned with Peachjar’s need to experiment with GenAI while maintaining long-term maintainability.
 

Solution Delivered

To jumpstart Peachjar’s GenAI journey, Stratus10 designed and deployed a solution using Amazon Bedrock Knowledge Bases, implemented entirely with Terraform. This approach ensures that the infrastructure is manageable, reproducible, and secure over time. The solution enables Peachjar's team to use natural language to ask questions and extract information from both flyer content and flyer engagement metrics.

Users can ask about available events, upcoming activities, or engagement metrics related to specific flyers or campaigns. By leveraging GenAI through Amazon Bedrock, Peachjar can begin realizing more value from its existing data while providing richer information to schools and communities.

The solution focuses on two distinct knowledge bases.
 

• Unstructured Knowledge Base

  The first knowledge base is an unstructured RAG-based knowledge base that uses Amazon OpenSearch Serverless as the vector database. Flyers are static PDF or image files stored in Amazon S3. In this implementation, flyers are not split into smaller chunks; instead, each file is treated as a single chunk to preserve document-level context.

  A multimodal foundation model is used to parse flyer files and extract text for ingestion. To ensure the knowledge base stays current, a scheduled AWS Lambda function scans the flyer storage location and ingests new documents if they have not already been processed.
   

• Structured Knowledge Base

  The second knowledge base is a structured knowledge base integrated with an Amazon Redshift cluster that stores flyer metrics. This setup enables natural-language Text-to-SQL generation, allowing queries to be automatically generated and executed against Redshift.

  Amazon Secrets Manager is used to securely manage Redshift credentials and control access to the database.
   

Additionally, IAM roles, IAM policies, and S3 bucket policies are applied throughout the solution to enforce granular permissions and maintain the principle of least privilege across all components.
 

Results and Benefits

Using GenAI enables non-technical users, as well as end users such as school administrators, to query complex datasets using natural language without introducing technical barriers or long turnaround times. This improves Peachjar’s ability to deliver value from both flyer content and engagement metrics and supports more timely access to information.

The solution also establishes a foundational GenAI capability within Peachjar. With this groundwork in place, Peachjar is positioned to continue experimenting and building on GenAI use cases, supporting future development and enabling deeper insight into how schools and communities engage with digital communications.
 

About Stratus10

Stratus10 is an AWS consulting partner specializing in cloud architecture, data, and GenAI solutions. We help organizations design, build, and optimize secure, scalable systems on AWS, with a focus on practical implementations that deliver measurable value.

 

Use case: GenAI
Client: Peachjar
Date: December 2025
Category: AI