Morsia Guitdam Hinimdou

SmartHub is an educational platform designed to modernize teaching and streamline communication between students, teachers, and administrators.

The solution integrates multi-agent chatbot assistants, automatic AI quiz generation, intelligent analysis of educational documents, and automation of academic processes.

🎯 Project Objectives

Centralize pedagogical activities (courses, quizzes, projects, internships, resources)

Automate AI-driven quiz generation and grading

Provide an intelligent pedagogical assistant based on LLM + RAG

Develop a modern, secure, and scalable web platform

🚀 Key Features

User Management (STUDENT, TEACHER, ADMIN)

Authentication and profile management

Course, announcement, and resource management

Project and internship management

Quiz creation, editing, and taking

Adaptive quizzes with detailed statistics

Automatic AI quiz generation

Role-based personalized dashboard

Multi-agent educational chatbot

Document-based RAG (PDFs, course materials, pedagogical resources)

Collaborative agents and supervision agents

Technical Architecture

Frontend

React + TypeScript

Backend

REST API developed with Spring Boot

Database

PostgreSQL

AI Layer

LLM · RAG · Multi-agent systems

Backend – SmartHub REST API

Technical Stack:

Java 21, Spring Boot 3.4.11, Spring Data JPA, PostgreSQL 13, Maven, Lombok

Layered Architecture:

Controller → Service → Repository → PostgreSQL

DTO usage for API contract isolation

JPA entities with constraints

Transactions via @Transactional

Enum Role persisted as STRING

Logging via Lombok @Slf4j

🔗 Links

Deployed Platform:

Full Project Documentation:

The main objective of this project is to predict the success or failure of startups based on historical data from Crunchbase (funding rounds, acquisitions, IPOs, industry sector, location, and other key indicators).

The work includes the entire Data Science pipeline:

✔ Data cleaning and preprocessing

✔ Feature engineering

✔ Handling class imbalance (SMOTE, Borderline-SMOTE, SMOTEENN, TomekLinks)

✔ Training and comparison of several Machine Learning models

✔ Performance evaluation

✔ Model deployment via a Flask API hosted on Render

🔹 Technologies Used

✔ Python 3.12

✔ Pandas

✔ NumPy

✔ Scikit-learn

✔ LightGBM

✔ XGBoost

✔ ExtraTreesClassifier

✔ SVM

✔ Matplotlib

✔ Seaborn

✔ Jupyter Notebook

✔ VS Code

✔ Google Colab

✔ Flask

✔ Gunicorn

✔ Render

✔ You can find all the necessary project elements here:

As part of this project, we developed an intelligent surveillance system to monitor people's behavior using image and video processing. Several datasets were collected from specialized fall detection websites and Kaggle to train our models.

Implemented Models

CNN: analyzes the spatial evolution of images.

LSTM: captures temporal evolution for fall prediction.

YOLOv8: real-time fall detection.

SVM & Random Forest: detection of fatigue and drowsiness.

Application

An interactive application was created with Streamlit, offering real-time monitoring based on computer vision to detect falls, drowsiness, and predict imminent falls, with the aim of improving people's safety and quality of life.