Sela. | Cloud Better.

Edusoft

Problem Statement/Definition

Edusoft Ltd., a subsidiary of ETS and a leader in technology-based English Language Learning and Assessment Solutions, faced the challenge of providing timely and personalized feedback to students on their English writing assignments. With an increasing demand for more effective language learning tools, Edusoft needed a system that could automatically evaluate writing assignments while allowing teachers to review and edit the feedback. To meet these demands, Edusoft sought an advanced AI-driven solution that could deliver high-quality feedback efficiently, while ensuring scalability, data privacy, and seamless integration with existing infrastructure.

Proposed Solution & Architecture

To address this challenges Sela had designed and implemented a GenAI-based Educational Feedback System for Edusoft, focused on providing immediate, reliable feedback to students while empowering teachers to refine and personalize the responses. This innovative solution utilized a combination of AWS services to achieve scalability, security, and integration with Edusoft's existing systems.

 

AWS Services Utilized:

  1. Amazon Bedrock: Central to the AI model development, Bedrock facilitated the implementation of multi-shot prompting and Retrieval-Augmented Generation (RAG) systems, enabling sophisticated and context-aware writing evaluations.
  2. Amazon OpenSearch: Deployed to enhance the system's search and embedding capabilities, ensuring efficient data retrieval and processing.
  3. Amazon Aurora MySQL: Used for storing metadata and application data, providing a reliable and scalable database solution.
  4. Amazon EKS (Elastic Kubernetes Service): Enabled the secure and scalable deployment of the AI-based evaluation system, ensuring seamless integration with Edusoft’s existing infrastructure.

 

The system was designed with content management guardrails to ensure that the AI-generated feedback adhered to appropriate standards, safeguarding against unwanted content.

 

Metrics for Success

Edusoft’s new AI-based Educational Feedback System provided great results among the pilot group:

  • Accuracy of Feedback: The AI system provided accurate and contextually relevant feedback in 90% of the test cases, meeting the project’s stringent quality benchmarks.
  • Teacher Efficiency: All teachers noted a substantial decrease in the time needed to give scores and feedback, attributed to the system’s capabilities.
  • System Integration: The solution was successfully integrated with Edusoft’s existing infrastructure, with no significant disruptions or compatibility issues reported.
  • User Satisfaction: Initial testing showed overwhelmingly positive feedback from both students and teachers, with 87% of users expressing satisfaction with the new system’s performance and ease of use.

 

Lessons Learned/Outcomes

 

The implementation of this AI-based system provided Sela and Edusoft with valuable insights:

  1. Customization is Key: Fine-tuning the AI-generated feedback prompts based on specific contexts and levels was essential in earning the trust and satisfaction of both teachers and students.
  2. Data Privacy: Implementing content management guardrails and local deployment within Edusoft's AWS cloud ensured that the solution met high standards of data privacy and security.
  3. Scalable Solutions: The project underscored the importance of scalable AI solutions, particularly in educational environments where demand can vary significantly.

 

In conclusion, Edusoft’s Educational Feedback System exemplifies the successful application of GenAI and its ability to address complex educational challenges, enhancements of the learning experience while maintaining high standards of security and efficiency.