01 / 20
Graduation Project · 2026

EduMapAI

Understand what you know. Discover what you don't.

An AI-powered learning platform that connects course material, quiz performance and AI analysis to reveal weak concepts, their likely causes, and what to study next.

Scroll to begin
02 — The Problem

More content than ever. Less clarity than ever.

Students rarely know what they actually understand — until the exam tells them.

Fragmented Material

Slides, PDFs, notes and videos scattered across sources with no connection between them.

False Confidence

Finishing a chapter feels like understanding it. Often, it isn't.

Generic Help

AI chat answers are disconnected from the actual course — and never explain why a student is weak.

03 — The Idea

One system that connects learning to understanding.

EduMap AI links the student's material, assessment results and AI analysis into a single loop.

01
Course Material
Documents uploaded per course
02
Learning Data
Concepts, questions, attempts
03
AI Analysis
Grounded in the material
04
Weakness Detection
At the concept level
05
Recommendations
What to study next
04 — Why AI?

Rules can count wrong answers. They can't explain them.

Traditional platforms report scores. Understanding which concept failed and why requires reading unstructured material and reasoning across questions — the work AI is actually good at.

Understand
course material
Connect
questions with course concepts
Analyze
quiz attempts
Detect
weak concepts
Identify
possible root causes
Recommend
what to study next
05 — How the System Works

From a student's click to a learning insight.

01
Student
Studies, uploads, takes quizzes
02
Frontend
Web interface
03
.NET Backend
REST API · control
04
AI Processing
Gemini + RAG
05
SQL Server
Persisted by backend
06
Learning Insights
Back to the student
06 — System Architecture

The backend is in control.

Every request passes through the .NET backend. It owns business rules, data flow and persistence.

The AI never touches the database directly.
It receives context from the backend and returns results through defined contracts. The backend stores them.
Frontend
Web client consuming the APIs
.NET Backend
REST API entry point
Application / Business Logic
Rules, validation, orchestration
AI Integration Layer
Gemini · RAG · analysis
SQL Server
Single source of truth
backend controlled
07 — Frontend → Backend → AI → Database

Four layers. Clear responsibilities.

L1
Frontend
  • User interaction
  • Course interface
  • Documents
  • Quizzes
  • Results
  • Learning insights
L2
Backend
  • Authentication
  • Authorization
  • API endpoints
  • Business logic
  • Data validation
  • Persistence
  • AI orchestration
L3
AI
  • Document understanding
  • Retrieval
  • Quiz analysis
  • Weakness detection
  • Root-cause analysis
  • Recommendations
L4
Database
  • Users
  • Courses
  • Documents
  • Concepts
  • Quiz attempts
  • Results
  • Learning data
  • AI-generated insights
08 — API Contracts

A controlled boundary between every layer.

The frontend talks to the backend only through defined REST contracts. The backend decides what reaches the AI — and what gets stored.

frontend → .NET REST API (examples)
POST/api/quiz-attemptsSubmit a quiz attempt
GET/api/courses/{courseId}Load a course
GET/api/learning-insights/{userId}Fetch insights
09 — user_id and course_id

Whose data is this? The backend decides.

user_id
course_id
quiz_attempt_id
.NET BACKEND IDENTIFIES
  • Which student
  • Which course
  • Which assessment
  • Which concepts
AI receives structured context
Only the relevant, scoped data — never free access to the database.
10 — Quiz Attempt Analysis

One attempt. Eight steps to an insight.

01 · Frontend
Student answers quiz
02 · Backend
Backend records attempt
03 · Backend
Answers mapped to concepts
04 · AI
AI analyzes performance
05 · AI
Weak concepts detected
06 · AI
Possible root causes identified
07 · AI
Recommendations generated
08 · Backend
Backend stores the insights
11 — Document Processing

Uploaded material becomes searchable knowledge.

A PDF is not knowledge yet. EduMap AI extracts its structure, breaks it into chapters and concepts, and indexes it so the AI can retrieve exactly what's relevant.

01Document
02Extraction
03Text / Structure
04Chapters
01Concepts
02Relationships / Keywords
03Searchable Knowledge
04AI / RAG
12 — Gemini + RAG

Answers grounded in the student's own course.

Retrieval-Augmented Generation fetches relevant passages from the course material before Gemini responds.

Without RAG
Generic model knowledge
With RAG
Grounded in actual material
01Student question / quiz result
02Retrieve relevant course content
03Provide context to Gemini
04Generate grounded analysis
05Return result through backend
13 — Weakness Detection

Not "weak at Chapter 3."
Weak at this concept.

LEVEL 1
Course
LEVEL 2
Chapter
LEVEL 3
Concept
LEVEL 4
Questions
LEVEL 5
Student Answers
LEVEL 6
Performance Pattern
DETECTED
Weak Concept
14 — Root Cause Analysis

Detecting a weakness isn't the same as understanding it.

01Low score
02Weak concept detected
03Analyze related questions
04Compare with prerequisite concepts
05Identify possible root cause
Likely causes the system distinguishes
Lack of understanding
Missing prerequisite knowledge
Repeated misconception
Insufficient practice

An inference — a likely cause, not a guaranteed diagnosis.

15 — Personalized Recommendations

From analysis to a concrete next step.

01
Weak Concept
02
Root Cause
03
Relevant Course Material
04
Recommended Study Action
→ Review a specific concept
→ Revisit a prerequisite
→ Read a relevant section
→ Practice related questions
16 — The Team

Seven people. One system.

AM
Ahmed Mohamed
Team Leader — Backend
ME
Mohamed Essam
Frontend Developer
MB
Mohamed Bayommi
Frontend Developer
MA
Mohamed Ali
DevOps Engineer
MS
Mohamed Samir
DevOps Engineer
FM
Fatma Mahmoud
AI Engineer
HM
Hamdy Mohamed
UI/UX Designer
17 — Current Status · Core System

What the core system covers today.

Course material processing
Documents
Concepts
Quiz / assessment analysis
AI integration
RAG
Weakness detection
Root-cause analysis
Recommendations
Backend API
SQL Server persistence
Core · current scope
18 — Future Vision · Planned

Where EduMap AI could go next.

Planned developments — not part of the current implementation.

Planned
Study Planner

Turn recommendations into a scheduled study path.

Planned
Gamification

Motivation through progress and achievements.

Planned
Study Communities

Learn together around shared courses.

19 — Limitations

Honest about what AI can — and can't — do.

AI output is probabilistic.
Root-cause detection is an inference, not a guaranteed diagnosis.
Large documents require careful processing and retrieval.
AI quality depends on the quality and structure of the learning material.
RAG improves grounding but does not guarantee perfect answers.
Processing large documents can require extra time and compute.
20 — Closing

EduMap AI turns learning data into understanding.

Thank you · Questions welcome