# 🧠 AI Tutor V2 ---- Concept Note <br><br> ### 1. Project Title AI Tutor for ABS Elevate e-learning. <br> ### 2. Purpose Enhance student understanding by providing an AI-powered chatbot that delivers instant, personalized support based on the course content. <br> ### 3. Target Users • Students • Teachers (for monitoring or support) <br> ### 4. Core Features ``` • In-app Chatbot: Embedded directly in the e-learning platform. • Course-Aware: AI Trained on lessons, quizzes, and media content. • Natural Q&A: Understands and answers questions in plain language. • Contextual Memory: Supports follow-up questions using chat history. • Real-Time Page Context: Scrapes and uses current page content to answer “Explain this” type queries. • Concept Reinforcement: Helps clarify and deepen understanding of key topics. • Interaction Logs: Generates usage data and insights for reporting. ``` <br> ### 5. Tech Stack • Frontend: JavaScript chatbot embedded directly into the e-learning. • AI: Custom RAG pipeline with contextual query understanding, chunk retrieval, and reranking • Memory: Redis for session and chat memory management • Embedding & Vector Store: Advanced data pipeline to vectorize diverse content types; stored in Supabase. • Backend: Python and Node.js services for core logic and API handling • Orchestration: n8n for managing session flow, memory, and system coordination <br> ### 6. Generalization & Scalability What makes this AI Tutor stand out is its high scalability with minimal overhead. The system can be deployed across multiple courses regardless of content type, structure, or language — with virtually no additional cost or effort. > We’ve successfully tested it with complex setups, including SCORM packages embedded within other SCORMs, interactive games, and diverse media types. The tutor managed to understand and explain even gamified content in real time (see Live Examples), demonstrating its adaptability across various learning formats. This positions the tool as a plug-and-play enhancement for any course across departments or subjects. <br> ### 7. Data & Metrics Tracking We’ve integrated a robust user authentication and tracking system, enabling the collection of consolidated KPIs across all AI Tutor instances. This includes: ``` • Number of users per course • Number of questions asked • Type of questions asked • Engagement rate per module • Frequency and duration of interactions • Student-level insights for educators and admins ``` These metrics allow us to monitor usage, identify gaps in content comprehension, and continually improve both the tutor and the learning experience. All data is stored and structured to support future reporting, dashboarding, and learning analytics. <br> ### 8. Recommendation for Large-Scale Adoption The AI Tutor is ready for large-scale deployment. Beyond improving understanding and engagement, it also: ``` • Bridges accessibility gaps (e.g., multilingual support, instant help) • Reduces pressure on subject-matter experts, who may not be immediately available • Enables consistent support across all content types and time zones ``` Scaling this tool organization-wide will not only enhance learner outcomes, but also reduce repetitive teacher support, improve data-driven decision-making, and future-proof the e-learning experience. <br> ### 9. Success Criteria • Improved course completion rates • Student satisfaction scores • Reduced teacher intervention on FAQs <br> ### 10. Live Examples 🎬 1. The bot identified the current lesson and explained it to the student in a summarized, easy-to-understand way. ![image](https://hackmd.io/_uploads/BkC7gE8Oxg.png) 2. The tutor breaks down specific parts of the content and helps the user craft the best possible response. ![image](https://hackmd.io/_uploads/H1Nox4Udel.png) 3. No more worries about explaining gamified content — that’s now the Tutor’s job. The AI Tutor can clearly walk users through interactive activities like the AI Terminology game, explaining how to play and guiding them step by step when needed. ![image](https://hackmd.io/_uploads/B1s9ZVLOxg.png) 4. **Accessibility**: The bot can understand and respond in multiple languages. As shown below, it accurately understood a question in **Darija** and replied in the same dialect. ![image](https://hackmd.io/_uploads/HyOl4VUdxx.png) <br> <br> <br> ### 11. Summary This is just the tip of the iceberg — the AI Tutor is now capable of answering a wide range of questions, making e-learning more engaging, accessible, and enjoyable for students of all levels.