The SUCCEEDU web application is a comprehensive platform designed to empower educators by streamlining resource creation and automating tasks. Leveraging AWS services, the solution will deliver a scalable, secure, and efficient infrastructure, incorporating AI-powered tools for question generation, adaptive learning paths, and automated lesson planning.
The Challenge
Teachers today face mounting workloads, growing administrative demands, and increasing pressure to personalise learning for diverse student needs. Resource creation, including practice sets and question materials, is largely manual and time-intensive, reducing the time available for student engagement and instructional innovation. Existing materials are often generic, outdated, and inflexible, making it difficult to adapt resources to specific classroom requirements or curriculum updates. The absence of a unified, creative, and curriculum-aligned solution forces teachers to rely on fragmented tools and platforms, further compounding inefficiencies and limiting their capacity to focus on teaching excellence.
The Solution
At the core of the solution is Amazon Bedrock, which powers a GenAI Bedrock Agent capable of interpreting user prompts and returning personalised educational resources. These include AI-generated lesson plans, teaching materials, curriculum-aligned content, and dynamic visual assets. The Bedrock Agent orchestrates interactions with backend AWS services to process requests, enrich prompts with educational context, and retrieve or generate the most relevant outputs.
The platform’s frontend is delivered using Amazon S3 for static assets combined with Amazon CloudFront for global low-latency content distribution. AWS WAF protects the application from common web-based threats such as bot traffic and malicious requests, while Amazon Route 53 provides highly available DNS resolution and intelligent traffic routing.
When users interact with the SucceedU application, requests are securely routed through Amazon API Gateway to AWS Lambda functions that execute serverless backend logic. These functions coordinate with Amazon Aurora (RDS) for storing and querying structured educational content metadata, Amazon DynamoDB for high-speed access to lesson templates, personalisation profiles, and real-time user interactions, and Amazon S3 for durable storage of generated learning assets such as documents, images, and multimedia resources.
All AI inputs and outputs, including instructional prompts, content generation queries, and personalisation requests, are exchanged with the Bedrock Agent. The agent returns high-quality, context-aware educational content that is formatted and delivered to end users in near real time, ensuring responsive classroom-ready experiences.
In Phase 2, the platform introduces an advanced AI-powered image generation pipeline to significantly improve the quality, accuracy, and educational relevance of generated visual content. This enhancement enables the Bedrock Agent to generate high-resolution illustrations, diagrams, infographics, and subject-specific classroom visuals aligned with curriculum objectives and age-appropriate learning standards.
When visual assets are requested, prompts are processed through an AI orchestration layer that applies optimised prompt engineering, visual style conditioning, and curriculum metadata enrichment before invoking image generation models. Generated images are automatically post-processed to optimise resolution, improve clarity, ensure accessibility compliance, and apply optional branding or watermarking requirements.
Generated visual assets are securely stored in Amazon S3 and indexed within DynamoDB and Aurora to enable fast retrieval, versioning, reuse, and personalised recommendations. Frequently accessed visuals are cached and globally distributed through CloudFront, ensuring consistent low-latency delivery to educators and students worldwide.
A continuous improvement feedback loop is also introduced in this phase. User engagement data, content ratings, and usage analytics are collected to refine prompt templates and generation parameters over time, enabling progressive improvements in visual quality, relevance, and instructional effectiveness.Operational visibility and governance are provided through a comprehensive observability layer. Amazon CloudWatch delivers real-time system metrics and performance monitoring, while Amazon CloudTrail captures audit logs for security and compliance tracking. Amazon SNS and Amazon SES are used for automated notifications and alerts, including content generation completion updates, system health alerts, and operational error reporting. Together, this architecture delivers a scalable, secure, and AI-optimised education platform that combines intelligent content generation with high-quality visual learning assets, enabling SucceedU to deliver personalised, engaging, and future-ready digital education experiences.
The Results
With the introduction of AI-powered visual generation in Phase 2, educators gained access to on-demand, curriculum-aligned visual assets including diagrams, infographics, subject illustrations, and classroom-ready visuals. This resulted in:
- 65% reduction in time spent sourcing external teaching visuals
- 40% increase in lesson material reuse due to centralised storage and indexing of generated visuals
- Improved classroom engagement metrics based on higher student interaction with visual learning materials
Educators reported higher satisfaction with visual consistency, age-appropriate design quality, and improved alignment between lesson content and supporting imagery.
Lessons Learned
- Phase 2 demonstrated that high-quality visual generation required more than raw prompt input. Introducing curriculum metadata enrichment, structured prompt templates, and visual style conditioning significantly improved output relevance and reduced regeneration cycles.
- Storing large volumes of AI-generated content emphasised the importance of aligning storage tiers with real user behavior. Frequently reused templates and visuals benefited from low-latency storage and CDN caching, while historical drafts performed better in lower-cost archival tiers.
