Edu Agent

An intelligent tutor that learns how you learn — adapting its depth, tone, and focus based on your mastery, your history, and exactly where you got stuck last time.

scroll
10+
Modules
3
Difficulty Levels
360°
Learning Loop
Adaptive Memory

Project Resources

Everything, one
click away.

How It Works

The adaptive
learning loop.

01
Question
Learner asks an AI/ML question in chat
02
Classify
Model predicts Beginner / Intermediate / Advanced
03
Detect Topic
Cached index identifies the AI/ML concept
04
Retrieve
Relevant examples fetched from dataset
05
Tutor
LLM explains at the right level with memory context
06
Follow-up
Evaluator generates a conceptual check question
07
Evaluate
Response scored: good / partial / poor understanding
08
Memory
Mastery, weak spots, history saved to database
09
Dashboard
Learner snapshot and progress charts refresh
10
Adapt
Tutor evolves across every future session

About the Project

More than a chatbot.
A learning system.

EduAgent is a complete adaptive learning loop — not just a wrapper around an LLM. It classifies learner difficulty with a fine-tuned transformer, detects the AI/ML topic in real time, retrieves supporting examples, and prompts the tutor with the learner's full history.

What makes it genuinely adaptive is the Memory Agent. It tracks mastery scores, weak concepts, topic revisit frequency, and the last evaluator result — injecting all of this into every future prompt. The tutor literally changes how it teaches based on where you struggled.

The Evaluator Agent generates a follow-up question after each explanation, scores the response with validated structured output, and feeds that signal back into memory — closing the loop between teaching and assessment.

Adaptive DifficultyFine-tuned classifier predicts level with confidence calibration and narrow intent override
Persistent Learner MemoryMastery, weak areas, and topic history saved across all sessions in a local database
Evaluator Feedback LoopFollow-up Q&A scored with structured, validated output — understanding level and recommendations
Dataset-Backed RetrievalCached indexes retrieve relevant examples per level and topic without re-fitting on every query
Full Auth & SessionsLogin, signup, logout with secure password hashing and per-user profile persistence