Case Studies

Read how Integra helped Diglossia with AWS Generative AI solutions which helped improve student outcomes and measures literacy progress over time.

Since its inception in 2014, Diglossia has embarked on a journey fueled by a singular objective: to revolutionize Arabic literacy education. Driven by visionary leaders, we have committed ourselves to incorporating contemporary assessment data, cutting-edge technologies, and profound insights into the very framework of Arabic language education. 

Diglossia holds the exclusive rights to the TALA [Test for Arabic Language Assessment] and Mubakkir assessment platforms, that have been developed in accordance with the Arabic reading level standards established by Dr. Hanada Taha, Endowed Chair Professor of Arabic Language at Zayed University, utilizing state-of-the-art technology within a secure online ecosystem. All test materials are intricately crafted and developed by our dedicated team and regional Arabic language experts.

Diglossia GenAI Dubai Case Study
Diglossia GenAI Dubai Case Study

Diglossia stands out as the premier provider of a comprehensive program that encompasses assessment, training, strategic planning, data analysis, and reporting across all grade levels from KG2 to grade 11. Our research-based, valid, and reliable Arabic assessments are widely adopted throughout the region. 

These assessments play a crucial role in identifying student learning gaps, and monitoring progress in early literacy skills, including phonemic and phonological awareness, alphabetic understanding, oral reading fluency, and proficiency in reading, writing, and vocabulary at both elementary and secondary levels.

Diglossia’s standards-based Arabic assessments provide actionable data to target specific skills, improve student outcomes, and measure progress over time. Diagnostic data allows teachers to provide meaningful formative feedback and develop personalized, evidence-based intervention strategies.

The Challenge

Diglossia faced a critical challenge in scaling its ability to provide teachers with meaningful and actionable insights from complex student datasets. Traditionally, teachers relied on manual data analysis, which was both time-intensive and prone to human error. This not only slowed down the decision-making process but also reduced the ability to identify key trends and areas for intervention effectively. The lack of an AI-powered solution limited the platform’s ability to handle real-time analysis, complex queries, and multi-parameter insights, such as comparisons across subjects or tracking student progress over time.

 

Risks for Diglossia

Time Inefficiency: Teachers spending significant time manually analyzing datasets, which could otherwise be devoted to educational activities.

Missed Opportunities for Intervention: The inability to quickly identify patterns or trends in student performance data limited proactive educational interventions.

Competitive Disadvantage: As educational platforms globally began to adopt AI-driven analytics, Diglossia risked falling behind in its competitive landscape.

The Solution

To address Diglossia’s challenges, Integra developed a tailored AI-powered solution leveraging Amazon Web Services (AWS) and Generative AI technologies. The solution was designed to automate and enhance the process of querying student datasets, providing educators with real-time insights into student performance, attendance and trends.

Key Components of the Solution

AI Agent for Dataset Querying: A Retrieval-Augmented Generation (RAG)-based agent powered by Claude Haiku 3.0, hosted on Amazon Bedrock, was integrated to enable natural language querying of datasets. The AI agent provided instant answers to teacher queries. The backend was developed using FastAPI and deployed on AWS ECS Fargate for scalability and reliability.

Data Management and Security: Uploaded datasets are stored temporarily in Amazon S3, ensuring secure and compliant data handling. All data was encrypted using AWS Key Management Service (KMS) to meet regional compliance standards like GDPR.

Enhanced Multilingual Support: The AI assistant was configured to support multiple languages, including English and Arabic to accommodate Diglossia’s diverse user base.
Multilingual examples were integrated into the AI’s prompts using LangChain for better contextual understanding.

Monitoring and Optimization: LangFuse was implemented to monitor AI-agent performance and trace user interactions, providing valuable insights into system usage and areas for improvement. Performance metrics, such as query resolution time and user satisfaction, were tracked using Amazon CloudWatch.

Guardrails for Ethical AI: NVIDIA NeMo Guardrails were deployed to ensure the AI-generated responses were contextually accurate, relevant, and aligned with educational standards. The guardrails also filtered inappropriate queries or content to maintain ethical use.

Support Services Provided by Integra

Pre-Implementation: System architecture design, prompt engineering strategies, and multilingual capability development.


Post-Implementation: Continuous performance monitoring, updates to prompts and features based on user feedback, and security reviews to ensure data privacy.

The Result

The AI-powered solution delivered significant improvements in data analysis and usability for Diglossia’s educators, addressing the challenges of manual analysis and slow decision-making.

Key Metrics and Benefits:


Time Efficiency: Automated query processing reduced the time spent analyzing student datasets by 40%, allowing teachers to focus on teaching.


Proactive Interventions: Teachers reported a 25% improvement in identifying at-risk students and addressing performance gaps earlier in the academic term.