< about me />
Hello! My name is Ahmed Mrabet. I'm currently studying Artificial Intelligence Engineering at the National School of Computer Science and Systems Analysis (ENSIAS in French) . My obsession is to solve problems outside of the tech community with computer science concepts. My passion for AI spans across various domains, with a particular emphasis on Computer Vision and Reinforcement Deep Learning. I am driven by the desire to develop useful applications that make a positive and meaningful impact on people's lives. I thrive on working with diverse teams, sharing knowledge, and learning from each other to accomplish our goals together, so reach me out if you want to team up on any project . Beyond school, I love film photography , cuddling my cat , and enjoy some coffee.
< experience />
Machine Learning Engineer Intern @Yaakey, On-Site, Casablanca, Morocco
June 2023 – August 2023- Utilized high-resolution satellite imagery for automated building footprint extraction, replacing manual data collection methods.
- Developed a U-Net semantic segmentation model for accurate predictions.
- Benchmarked deep learning models to optimize footprint extraction accuracy.
- Built a polygonizer to convert segmentation into coordinates, significantly accelerating the collection of building footprint data.
- Integrated APIs for real-time data acquisition, enhancing model accuracy.
Skills Acquired
Geospatial Data · Semantic Segmentation · U-Net · Model Benchmarking · PyTorch · EfficientNet · API Integration · Geodataframes · Torchgeo · Torchvision
< projects />
This project implements an Arabic Information Retrieval system that enables users to perform search queries and retrieve relevant information in the Arabic language. The core components include text preprocessing, text vectorization, and text indexing, which form the backbone of any Information Retrieval system.
PythonNLTKTF-IDFFlaskJavascriptThis project implements a Hidden Markov Model (HMM) to model stock price movements. The model is trained using the Baum-Welch algorithm and makes predictions using the Viterbi algorithm. The model predicts whether the stock price will rise or fall in the following trading day.
PythonHMMViterbiBaum-WelchThis project aims to address the challenge of accurately predicting protein-protein interactions (PPIs) by leveraging the power of Graph Neural Networks (GNNs). PPIs are critical for understanding biological processes, and accurately predicting them is essential for unraveling the complexities of cellular mechanisms and designing effective therapeutics
TensorflowBiopythonGNNDesigning a simulation and the development and implementation of a smart traffic light controller using reinforcement learning. The use of reinforcement learning is motivated by the need to address the inefficiencies and problems caused by the existing fixed traffic light controllers.
PythonTensorflowDQNSumoThis project is a comparison of traditional backpropagation method and Cuckoo Search algorithm for training a neural network. It aims to find a better and efficient way of training a neural network by optimizing the weights and biases.
PythonCuckoo SearchNeural NetworkThis project implements a Minesweeper game using Python and Pygame. The game is played by a human player and a reinforcement learning agent. The agent learns to play the game by interacting with the environment and observing the rewards it receives.
PythonPygameDQN