NeuroMap – AI Knowledge Graph Builder for Personalized Learning

Microsoft AI Unlocked

---

NeuroMap is an AI-powered learning assistant that converts messy study notes into an interactive knowledge graph. Students upload lecture notes (PDF or text), and the system automatically extracts key concepts, maps relationships between them, and generates a structured visualization of the topic.

Instead of reading notes line-by-line, students can explore the concept network, identify weak areas through quizzes, and receive a personalized revision roadmap.

The goal of NeuroMap is to transform passive reading into visual, concept-driven learning.

What the App Does

NeuroMap helps students understand complex material faster by:

1. Parsing study notes

• Accepts uploaded lecture notes (PDF/text).

• Extracts readable text from documents.

2. AI Concept Extraction

• Uses an LLM to detect important concepts.

• Identifies relationships between them.

3. Concept Graph Generation

• Constructs a knowledge graph from extracted concepts.

• Displays it visually using an interactive node graph.

4. Interactive Concept Exploration

• Clicking a concept reveals explanation and quiz questions.

5. Weak Concept Detection

• Incorrect quiz answers mark concepts as weak.

6. Personalized Revision Plan

• Generates a roadmap based on weak concepts.

7. Personalized Revision Roadmap

• recommend study order, which topics require review

8. Learning Progress Track

• The system tracks concept mastery and displays overall learning progress to the user.

Why It Is Useful

Students often struggle with:

• Large volumes of lecture notes

• Identifying the most important concepts

• Understanding relationships between topics

• Knowing what to revise

NeuroMap solves this by converting notes into a visual knowledge map, allowing students to:

• Understand topic structure quickly

• Identify dependencies between concepts

• Focus revision on weak areas

• Learn interactively instead of passively reading notes

Tech Stack

Frontend

• React
• TypeScript
• Vite
• TailwindCSS
• ReactFlow (concept graph visualization)
• Axios

Backend

• Node.js
• Express.js
• Prisma ORM

Database

• PostgreSQL

AI / NLP

• Groq API (LLM)
• Prompt-based concept extraction
• Retrieval Augmented Generation (RAG)

Document Parsing

• pdf2json (PDF text extraction)

Performance Optimization

• Azure Blob Storage - file storage for uploaded notes
• Azure Redis Cache - caching AI-generated quiz and explaination responses

Deployment

• Backend: Render
• Frontend: Vercel

Project Structure

AIStudyBuddy
│
├── backend
│   ├── routes
│   │   ├── analyzeNote.js
│   │   ├── generateQuiz.js
│   │   ├── revisionRoadmap.js
│   │   └── studySession.js
│   │
│   ├── services
│   │   ├── conceptExtractor.js
│   │   ├── llmClient.js
│   │   ├── retrieval.js
│   │   ├── chunker.js
│   │   ├── blobStorage.js
│   │   └── cache.js
│   │
│   ├── db
│   │   └── prisma.js
│   │
│   └── server.js
│
├── frontend
│   ├── src
│   │   ├── components
│   │   │   ├── ConceptGraph.tsx
│   │   │   ├── ConceptPanel.tsx
│   │   │   └── RevisionPlan.tsx
│   │   │
│   │   ├── pages
│   │   │   └── UploadPage.tsx
│   │   │
│   │   └── App.tsx
│   │
│   └── api
│       └── api.ts
│
└── README.md

System Architecture

User
│
▼
React Frontend (Vercel)
│
▼
Express Backend API (Render)
│
├── Document Processing
│     └─ PDF Parsing (pdf2json)
│
├── AI Layer
│     ├─ Concept Extraction
│     └─ Explanation + Quiz Generation
│
├── Retrieval Layer
│     └─ Relevant Note Chunk Retrieval
│
├── Cache Layer
│     └─ Redis (Azure Cache for Redis)
│
├── Storage Layer
│     └─ Azure Blob Storage
│
▼
PostgreSQL Database
│
├── Notes
├── Concepts
├── Relationships
└── Quiz Results
│
▼
Knowledge Graph returned to frontend
│
▼
ReactFlow Visualization

Local Setup Instructions

1. Clone Repository

   git clone https://github.com/<your-username>/AIStudyBuddy.git
   cd AIStudyBuddy

2. Backend Setup

Navigate to backend:

cd backend

Install dependencies:

npm install

Create .env

DATABASE_URL=your_postgres_connection
GROQ_API_KEY=your_groq_api_key

AZURE_STORAGE_CONNECTION_STRING=your_blob_connection
AZURE_STORAGE_CONTAINER=notes

REDIS_URL=your_redis_connection

PORT=5000

Run server:

npm run dev

Backend runs on:

http://localhost:5000

3. Frontend Setup

Navigate to frontend:

cd frontend

Install dependencies:

npm install

Create .env

VITE_API_URL=http://localhost:5000

Start frontend:

npm run dev

Frontend runs on:

http://localhost:5173

API Endpoints

Method	Endpoint	Description
POST	/api/analyze	Uploads PDF/Text, returns nodes and edges.
POST	/api/quiz	Returns explanation and MCQs for a concept.
POST	/api/quiz/result	Saves score and updates concept strength.
GET	/api/roadmap/:noteId	Returns a personalized revision plan based on weak concepts.
GET	/api/study-session/:noteId	Returns the next concept the student should study.

Limitations and Challenges

During development several technical challenges emerged:

1. PDF Parsing Issues

Many PDFs contain irregular encodings, which caused text extraction errors. These were mitigated using safe decoding and text cleaning.

2. AI Rate Limits

Gemini free-tier API limits caused request failures during testing. Fallback logic was implemented to prevent crashes when AI responses fail.

3. Noisy Note Content

Lecture PDFs often include formatting artifacts that degrade concept extraction accuracy.

4. Graph Layout

Automatically organizing concept graphs in a readable layout required custom positioning logic.

Post-MVP Goals (Future Development)

Future improvements planned for NeuroMap include:

1. Better AI Knowledge Graph Extraction

Use structured extraction pipelines (RAG / embeddings) for higher concept accuracy.

2. Semantic Search

Allow students to search concepts directly inside notes using embeddings.

3. Multi-Document Knowledge Graphs

Merge multiple lectures into one connected knowledge map.

4. Smart Study Scheduling

Integrate spaced repetition to optimize revision timing.

5. Advanced Graph Visualization

Cluster related concepts and highlight core topics automatically.

6. Collaborative Learning

Allow students to share concept maps and annotations.

Conclusion

NeuroMap demonstrates how AI can transform traditional notes into an interactive learning experience. By combining document parsing, LLM-based concept extraction, and graph visualization, the system provides a powerful way for students to explore and revise complex subjects efficiently.

The current prototype validates the concept and lays the foundation for a scalable AI-driven learning platform.