Proposal
Slothub - Agentic Learning & Management Platform
(Optimizing teaching and learning through class management, AI-driven content, and personalized tutoring)
1. Executive Summary
Slothub is a comprehensive web-based ecosystem designed for teachers, students, and administrators to synchronize educational activities on a single platform. The platform manages classroom lifecycles, timetables, attendance, and assignments, while integrating a Multi-layered AI architecture (Slozy AI Tutor) to provide personalized learning roadmaps and RAG-based knowledge retrieval.
Our team of 5 IT students is developing Slothub using a modern microservice architecture deployed on AWS (ap-southeast-1) with a Multi-AZ design for high availability. The system integrates a React/TypeScript frontend, a Spring Boot backend, a FastAPI AI service, and AWS Bedrock AgentCore for the agentic tutor.
The goal is to reduce administrative burdens for educators while empowering students with guided, self-paced learning tools, leveraged by scalable AWS services such as Fargate, RDS PostgreSQL, and S3.
2. Problem Statement
What is the Problem?
- Fragmented Tools: Teachers must navigate multiple disconnected platforms (gradebooks, chat apps, files) for data management, leading to inefficiency.
- Lack of Personalization: Traditional e-learning offers a “one-size-fits-all” approach that fails to adapt to individual student knowledge gaps and learning paces.
- Administrative Overload: educators spend significant time on repetitive tasks like manual exercise creation and content compilation.
- Static Interactions: Existing solutions lack intelligent conversational interfaces capable of providing contextual guidance outside of classroom hours.
The Solution
Slothub centralizes educational data and workflows on a unified stack:
- Unified Management: Spring Boot API and RDS PostgreSQL for core operational data.
- AI-Generated Content: Dedicated containerized AI services calling OpenAI & AWS Bedrock for automated material generation stored on S3.
- Agentic Tutoring: Slozy AI Tutor (powered by Bedrock AgentCore) provides step-by-step guidance using RAG (Retrieval-Augmented Generation) from theory databases instead of generating hallucinatory content.
- High Availability: Multi-AZ deployment for both database and application layers via ALB and Fargate.
Benefits and ROI
- Centralization: All study tracks, classroom management, and assignments on one platform.
- Personalized Learning: Adaptive roadmaps, quizzes, and AI tutoring based on curriculum context.
- Scalability: Containerized microservices allowing independent scaling via ECS Fargate.
- Operational Reliability: Multi-AZ RDS and Fargate with monitoring via CloudWatch and security from GuardDuty.
- Cost Control: Optimized resource usage via Serverless execution (Fargate, Bedrock) and caching strategies.
3. Solution Architecture
Overview
User Access → Amazon Cognito (Auth) → WAF & ALB → ECS Fargate (Backend / AI Services) → RDS PostgreSQL (Multi-AZ) & S3 (Documents). The AI Assistant interacts with Amazon Bedrock AgentCore for intelligent tutoring logic.
AWS Services Used
| Service | Function |
|---|---|
| Amazon Cognito | User authentication and authorization |
| Amazon Route 53 | Domain routing and DNS management |
| AWS WAF | Web application firewall for threat protection |
| AWS Amplify | Frontend hosting and management |
| AWS Certificate Manager | SSL/TLS certificate management |
| Amazon VPC + IGW | Private networking with controlled internet access |
| Application Load Balancer | Traffic distribution to container services |
| Amazon ECS on Fargate | Serverless execution for Backend (Spring Boot) and AI (FastAPI) |
| Amazon ECR | Docker image repository for microservices |
| Amazon RDS (PostgreSQL) | Relational database with Multi-AZ Primary/Standby |
| Amazon S3 | Storage for documents, assignments, and uploads |
| Amazon Bedrock (AgentCore) | AI Tutor agent and generative language models |
| Amazon CloudWatch | Resource monitoring and logging |
| Amazon GuardDuty | Intelligent threat detection |
4. Technical Implementation
Implementation Phases
| Phase | Description | Suggested Duration |
|---|---|---|
| 1 | Infrastructure setup (VPC, ALB, Fargate, RDS Multi-AZ, S3, Cognito), IaC/CI basics | 2-3 weeks |
| 2 | RDS Schema design, Spring Boot API development, and Cognito integration | 3-4 weeks |
| 3 | AI Service (FastAPI) deployment, S3 document pipeline, Bedrock connectivity | 2-3 weeks |
| 4 | AgentCore (Slozy) implementation, tool binding to PostgreSQL and AI API | 2 weeks |
| 5 | Frontend production build, Amplify deployment, WAF/ACM/Route 53 configuration | 2 weeks |
| 6 | Load testing, CloudWatch observation, and cost/security tuning (GuardDuty) | 1-2 weeks |
Technical Requirements
- Client: Modern web browsers (Responsive interface).
- Backend: Java 17, Spring Boot 3.x, Spring Security (OAuth2/JWT).
- AI Services: Python 3.x, FastAPI, AWS SDK (Boto3) for S3 and Bedrock.
- Agent Core: LangGraph/LangChain, Bedrock AgentCore SDK.
- CI/CD: GitHub Actions for Docker build, ECR push, and ECS update.
5. Timeline & Milestones
| Milestone | Activity |
|---|---|
| Preparation | Finalize ERD/API contracts, User Roles, and Data Policies |
| Core Development | Classrooms, Assignments, Timetable, and Attendance modules |
| AI Integration | FastAPI + S3 + Bedrock; Slozy AI Assistant |
| Deployment | Production setup with ECR, Fargate, ALB, Amplify, and WAF |
| Operation | Monitoring, RDS backups, and security auditing |
6. Budget Estimation
Resource Investment Time
Total investment for the 5-member team contributing 20 hours/week over three phases:
- Phase 1: Platform Setup: 200 hours
- Phase 2: Core Features: 200 hours
- Phase 3: Completion & Deployment: 200 hours
- Total Estimated Time: 600 hours
Cloud Infrastructure Costs
- Hedged Costs: Utilizing the AWS Free Tier and credits package ($200) for development and initial deployment.
- Primary Items: Monitoring costs for RDS Multi-AZ, Fargate tasks, and Bedrock token usage via CloudWatch.
- Optimization: Rightsizing Fargate resources and implementing Bedrock caching.
7. Risk Assessment
| Risk | Impact | Probability | Mitigation |
|---|---|---|---|
| Data Sync Issues (Backend/AI) | High | Medium | API contracts, idempotent uploads, handled transactions |
| Bedrock Cost Overrun | High | Medium | Quota limits, result caching, optimized prompts |
| RDS Failover Latency | Medium | Low | Multi-AZ deployment, automated health checks |
| Security Breaches (API/S3) | High | Medium | IAM least privilege, WAF, Encryption at rest, GuardDuty |
| AI Output Quality | Medium | Medium | Prompt engineering, curriculum-restricted knowledge base |
8. Expected Outcomes
Success Criteria
- Functional MVP: Successfully manage classrooms, assignments, and AI-driven tutoring for the pilot group.
- Full Infrastructure: Secure deployment on AWS using ECS Fargate, RDS Multi-AZ, and Cognito.
- Performance: Acceptance latency for teacher/student interactions (monitored via CloudWatch).
- AI Accuracy: Slozy AI Tutor successfully retrieves valid theory via RAG for >90% of curriculum queries.
- Budget Control: Operates within the $200 AWS credit budget.
Long-term Value
- Extended Content: Automated Question Bank expansion and AI analytics for student progress.
- Scalability: Mobile application support and organization-wide multi-tenancy.
- Efficiency: Transitioning from administrative management to AI-augmented teaching.
9. References
- AWS Pricing Calculator
- Amazon Bedrock AgentCore and Amazon ECS Fargate Documents
- Project repository:
frontend,backend,AI,agent-corefolders.