Profile picture

Magdy Mahmoud

MSc Student at Technical University of Munich

magdy.mahmoud [at] tum.de

CV

About

I am a master student at TUM. I received my Bachelor's degree in Computer Science from Thebes University. Before starting my MSc at TUM, I have worked as a machine learning engineer at Aigorithm, tackling a variety of computer vision problems such as object detection and segmentation.

I've passion for problem solving. It was not until my undergraduate studies that I learned programming and quickly became obsessed with problem solving. My thinking skills has significantly improved through solving more than two thousands algorithmic problems since then. During years of training for problem solving, I've realized my research interests. I'm always looking to learn something new, to hone my problem-solving skills and to challenge myself in new ways.

Research

My long-term research goal is to make computers fully understand our life using computer vision. I appreciate the non-intrusive nature of computer vision systems. A computer vision system can merely observe a space and operate with no intervention or explicit input.

I am currently motivated by challenges in scene understanding such as Human-Object-interaction, visual tracking and Video Captioning as a mean for better scene parsing and understanding.

Projects

  • Road damage detection Pytorch, Python - Using Yolov5 for detecting road damages, fine-tuning and changing various parts of the code to achieve better result.
  • Human-Action-Recognition Keras, Python - Using keras for building models to predict human action in videos using ucf101 dataset.
  • Arabic-Image-Captioning Keras, Python - Used Flickr8k_Dataset and Bing translator for translating, Used Arabic word2vec, result is not good because the bad translation.
  • Stores Future Sales Prediciton Sklearn, Python - Dataset from kaggle competition Predict Future Sales competition. Time-series task, task is to predict sales of next month, Did feature engineering, Added mean encoding features and lagged features, Used an ensemble of RandomForest, lightgbm, SGD.
  • Gender-Age predicition Keras, Python - Given a face image, predict whether the person is male or female, Got a validation accuracy of 0.889 on Adience Benchmark dataset, Used a pretrained InceptionResNetV2 model, working on age predication using IMDB dataset.
  • JSP, Servlets, MySQL - This a questions and answers system (like ask.fm) with functionality such as ask/answer questions, search users, add friends, etc.
  • Online mobile store JSP, Servlets, MySQL - Ecommerce website: adding/view product/category..etc.
  • Education-management-system C++ - student register courses, adding/manage students/courses/assignment.
  • Tic tac toe c++ - console, Player vs. player/PC ( Using Minimax algorithm and Alpha–beta pruning).

Online Courses

  • Udacity Machine-Learning-Engineer-Nanodegree-Program [Certificate]
  • Deep Learning, a 5-course specialization by deeplearning.ai [Certificate]
  • Machine Learning, a 4-course specialization by University of Washington [Certificate]

Programming Competitions

Competitions Achievements (ICPCID)

Competitions Activity

  • Judge at 4 local ACM contests 2020
  • Tester at The 2019 Palestinian Collegiate Programming Contest
  • Judge at 3 local ACM contests 2019

My accounts on programming competitions websites:


Trajectory

  • Software Engineer Internship

    Google

    Aug 2022 - Dec 2022

    Working with the shopping ML team

  • MSc Informatics

    Technical University of Munich

    April 2022 - March 2024

  • Full-time Machine Learning Engineer

    Aigorithm

    Nov 2020 - April 2022

    Working on two problems: First, detecting road damage, training different models, such as detecron2 and yolov5, using different datasets such as rdd and our labeld data and testing different pipeline for processing images such as segmenting road from image before prediction. Second, Classifying face skin problems such as acne. Using our collected data and SD-260 dataset. Using face segmentation models and classical approaches

  • Computer Vision Internship

    Aigorithm

    Aug 2020 - Oct 2020

    Worked on building a generic framework for parallelizing preprocessing and loading multiple datasets at once using Python, TensorFlow-data and tensorpack. Fine-tuning yolov5 for custom datasets using Pytorch, Training on AWS using COCO and OpenImages, and building scripts for automating AWS usage using Boto3.HPC

  • Computer Vision Internship

    DevisionX

    Jan 2018 - Aug 2018

    A project to automatically verify ID information, I worked on developing models for segmentation and recognition of Arabic words using TensorFlow, and also did some backend development using python Flask

  • BSc Computer Science, GPA: 3.44/4

    Thebes University

    Oct 2014 - Aug 2018

    First in class in the first 2 years
    A in GP