About Customer:
Customer is a Chennai-based EdTech company building an agentic, generative-AI learning platform that personalizes education to each learner’s style and pace. The platform generates custom study materials, tutorials, quizzes, and real-time doubt resolution, alongside comic-style visual narratives, podcast-style audio, animated explainers, and short instructional videos generated directly from learning content.
Industry: EdTech — AI Learning
Offering: AWS Bedrock + Generative AI
Business Challenges:
- Manual server logins and ad-hoc deploy scripts made releases slow and inconsistent across every surface of the platform – RAG tutoring, comics, podcasts, notebooks, and video.
- Secrets were bundled with build artifacts or hardcoded in the application, creating security risk on every release and making key rotation difficult.
- No unified architecture existed for RAG tutoring, assessment generation, and video workflows. Scattered infrastructure slowed content generation and raised operating costs.
- Ad-hoc compute and storage for generative video workloads limited the scalability of animated and text-to-video learning content.
- Legacy and newer login modes coexisted without strict scoping, risking cross-account data bleed in saved learner progress.
Solution Implemented:
- An automated CodePipeline → CodeBuild → CodeDeploy release flow with zero-downtime Application Load Balancer traffic shifting, enabling daily, predictable releases.
- Deploy-time secret injection from a dedicated, IAM-controlled S3 bucket – credentials never touch source control or build artifacts.
- A unified AWS Bedrock architecture: S3 and Lambda-based embeddings feed an Amazon OpenSearch knowledge base, powering grounded, context-aware RAG tutoring.
- A GPU-backed Amazon ECS pipeline running the WAN 2.1 model, with Meta Llama 3 70B for prompt optimization, Nova Canvas for imagery, and Amazon Polly for narration.
- JWT-backed sessions with strict per-account data scoping in Amazon DynamoDB, keeping learner progress and preferences correctly isolated.
Services Used:
- Amazon Bedrock
- Lambda
- OpenSearch
- ECS / ECR
- SQS
- DynamoDB
- API Gateway
Business Benefits:
- 80% less release effort – automated CI/CD took releases from manual and error-prone to fast and predictable.
- Zero secret leaks – credentials are injected securely at deploy time, never stored in code.
- 85% target accuracy – AI tutoring delivers grounded, instruction-adherent answers for every learner.
- ~6-second, 480p video clips generated reliably – GPU-backed AI turns study content into video lessons at scale.
- Zero cross-account data bleed – every learner’s progress and data stay fully isolated.
Ready to Power Your EdTech Platform with Generative AI?
Click Here to Get a Free GenAI Readiness Assessment
Solutions Director – AI/ML • DA - Solutions, Rapyder Cloud Solutions
Frequently Asked Questions
Rapyder helped an EdTech company move from slow, manual, error-prone deployments to a fully automated, secure, and scalable CI/CD and GenAI architecture on AWS — cutting release effort by 80% while eliminating secret leaks and data isolation risks.
By building an automated release pipeline using AWS CodePipeline, CodeBuild, and CodeDeploy, combined with zero-downtime traffic shifting through an Application Load Balancer — replacing manual server logins and ad-hoc scripts.
80% reduction in release effort, zero secret leaks, 85% tutoring accuracy, reliable AI video generation at scale, and zero cross-account data bleed.
Depending on the use case, organizations can benefit from:
- Faster and more reliable software releases
- Improved application security
- Scalable AI infrastructure
- Better AI response quality
- Secure multi-tenant architecture
- Lower operational overhead through automation
While this case study focuses on an AI-powered EdTech platform, Rapyder's Generative AI and cloud modernization capabilities can be applied across industries such as BFSI, Healthcare, Manufacturing, Retail, Media & Entertainment, Logistics, and Enterprise SaaS. (This last point reflects Rapyder's broader capabilities rather than the case study itself.)
