Peter Antoine Harfouche

Experience

• 

  Saint Joseph University of Beirut

  Jul 2018 - Aug 2018

  Intern

  Development of an Electronic Circuit In a collaborative effort; The team skillfully created an electronic circuit integrated with a DHT22 sensor for monitoring room temperature and humidity. The circuit was seamlessly connected to a Raspberry Pi 3.Data Visualization and Control Interface: Showcased programming acumen by developing a Python 3 code to graphically represent sensor data. The project advanced into the realm of IoT by enabling remote sensor control via the internet, utilizing an MQTT server for command transmission.Innovative iOS Application (partially completed): Culminating the project, the team successfully implemented an iOS application. This application not only allowed remote interaction with the sensor but also enabled efficient data visualization. Show less

• 

  Potech Consulting

  Apr 2019 - Jun 2019

  Intern

  My experience at Potech Consulting marked a pivotal moment in my career, steering me towards a deep interest in cybersecurity. I embraced the challenge of enhancing our data monitoring capabilities, diving into the world of open-source tools for data collection, storage, analysis, and visualization.Faced with the need to handle vast volumes of data, I initiated extensive research to find open-source solutions that could meet our goals. However, realizing the unique needs of our company, I proposed an innovative approach: developing a custom application tailored to our specific requirements.This venture led to the creation of an efficient, robust application capable of managing and processing thousands of data files within seconds. Not only was the application well-received internally, but it also marked our foray into contributing to the open-source community. The project is now available on GitHub, showcasing our commitment to innovation and collaboration in the cybersecurity field.Through this experience, I gained invaluable insights into practical cybersecurity applications, underscored by a keen interest in developing solutions that are both effective and accessible. Show less

• 

  IRD

  Jun 2021 - Jul 2021

  Software Engineer Intern

  Created an Android mobile application from the ground up, utilizing Android Studio and Java. This initiative demonstrated proficiency in mobile app development and user interface design.Pioneered the integration of an Arduino board connected to environmental sensors with the Android application. This involved the real-time collection and cloud transmission of environmental data, showcased through interactive graphs, demonstrating expertise in IoT and data visualization.

• 

  HES-SO Haute école spécialisée de Suisse occidentale

  Jul 2021 - Sept 2021

  Software Engineer Intern

  Spearheaded the transformation of over 10 web pages for the Travel Therapy Project, aligning them with modern design principles and the strategic vision of the MINT agency. This redesign significantly enhanced the visual appeal and user navigation experience.Developed a user-centric interface for the virtual travel project, aimed at providing elderly users with easy access to various travel destinations. Utilized Angular and Node.js to create an engaging and intuitive platform, seamlessly guiding users from exploration to transaction. Show less

• 

  HumanTech Institute

  Jan 2022 - May 2022

  Software Engineer • Final Year Project

  In our collaborative Final Year Project, we embarked on an innovative journey to create a proof-of-concept for a solar-powered smart home recommender system. This ambitious project aimed to significantly reduce grid energy consumption by 15%.Our teamwork was pivotal in creating a dataset, incorporating detailed meteorological data to effectively model user profiles, a necessary approach due to the lack of real-world data. We explored machine learning techniques by implementing algorithms capable of accurately predicting solar energy production and user behavior. These algorithms were meticulously designed to optimize energy usage while offering a deep understanding of user interactions. Show less

• 

  IDEMIA

  Oct 2022 - Jul 2023

  Projet Fil Rouge: Action Recognition in Videos Using Pedestrian Keypoints

  During this academic project at IDEMIA (team of 3 students), I engaged in an innovative project focused on developing a skeleton keypoint-based action detection system. This project, aimed at enhancing the accuracy of practical applications such as unattended luggage detection, allowed me to delve deep into the realm of advanced keypoint detection. Leveraging MMPOSE, I played a crucial role in training LSTM and 2sAGCN deep learning models. These models were meticulously trained on a self-curated dataset of 350 videos, each manually annotated with action sequences, ensuring the highest level of precision in our models.The cornerstone of our project was the rigorous application of evaluation metrics, which was instrumental in refining our models. This meticulous approach led to a significant improvement in the accuracy of threat detection systems, contributing immensely to reducing delays in public transport and elevating safety measures.Furthermore, I was responsible for engineering a comprehensive processing pipeline. This pipeline was meticulously designed, starting from precise keypoint extraction and culminating in model prediction, ensuring a synchronized display of results for immediate action recognition. The success of this project not only highlighted the potential of advanced machine learning techniques in real-world applications but also underscored my commitment to leveraging technology for enhancing public safety and efficiency. Show less

• 

  Enedis

  Jul 2023 - Jan 2024

  Machine Learning Engineer • AutoML Internship

  Optimized French NLP Models: Significantly enhanced CamemBERT and FlauBERT by implementing quantization and PEFT techniques, including QLoRA. Resulted in a substantial 20 GB reduction in memory usage and a 60% decrease in computing time.Advanced Model Benchmarking: Developed and fine-tuned automated processes for hyperparameter optimization. This led to a remarkable 92% improvement in benchmarking efficiency and predictive accuracy across various use-cases, as reflected in key metrics like accuracy, F1, precision, and recall.Benchmarking Tools Scalability: Played a pivotal role in enhancing the scalability and reliability of benchmarking tools. Achieved smoother workflow operations and efficient troubleshooting by integrating Airflow for orchestration and exploring Fully Sharded Data Parallelism (FSDP). Show less

• 

  WeatherMind

  Apr 2024 - Nov 2024

  Deep Learning Engineer