Ellerin - Multi-agent AI System for education
An educational application that combines robust full-stack architecture with guided AI interactions to help students learn topics, practice understanding, and verify progress.
Project description
Ellerin is built on the MERN stack and integrates the Claude SDK for AI-driven tutoring workflows. It is designed to move beyond simple chat by guiding students through a structured learning path from setup to assessment.
The repository is currently private while development is ongoing, but the product direction is centered on reliable backend orchestration, a sleek and practical frontend, and strong educational outcomes.
Core features
Comprehensive setup flow
Intake and configuration stage that narrows the context window before learning sessions begin.
Context-engineered learning agent
A guided chat experience tuned for teaching, not just answering, keeping interactions aligned with learning goals.
General student chat
Flexible topic discussion where the user can ask questions and explore concepts with an educational AI assistant.
End-of-session quiz
Quiz workflow that checks what the student learned and provides a concrete feedback loop.
Robust backend architecture
API and service design focused on stable AI session handling, maintainability, and clear domain boundaries.
Sleek, focused frontend UX
Frontend experience designed for clarity and flow so students can stay focused on learning tasks.
Technical stack
- Frontend: React with reusable UI components for learning flows.
- Backend: Node.js + Express for API and orchestration layers.
- Database: MongoDB for user, session, and progress data models.
- AI integration: Claude SDK for guided tutoring behaviors.
- Full-stack structure designed for iterative feature growth.
Learning design approach
- Start with structured setup to improve context quality.
- Use guided chat for concept teaching and active exploration.
- Finish with a quiz to validate understanding.
- Keep the learning loop clear: explain, practice, verify.
What I learned
This project taught me how to treat AI as one subsystem inside a larger product architecture. The hardest part was not building chat itself, but designing reliable context preparation, user flow control, and meaningful evaluation of learning outcomes. It strengthened both my backend design skills and my product thinking.