Abdelwakil Benabdi

Play with our tools to make your path 💫: - Crawl2File (bêta): Crawl2File is a data extraction platform that enables users to transform unstructured web pages or documents into clean, structured datasets in seconds. The platform allows users to crawl websites or upload files and export the extracted data in formats such as JSON, CSV, Markdown, or SQL. - DeepSearchJobs: DeepSearchJobs is a job discovery platform designed to uncover hidden job opportunities that traditional job boards miss. Instead of relying on saturated platforms where thousands of candidates apply to the same listings, DeepSearchJobs continuously analyzes company websites and online sources to surface rare, early-stage, and low-competition opportunities.

Mathematics & Computer Science Exam preparation simplified with AI.

For all your Data, Business Intelligence, Real Estate ERPs and Web Apps projects, do not hesitate to contact me at this email address: [email protected] Pour tous vos projets Data, Business Intelligence, ERPs Immobilier et Web Apps, n'hésitez pas à me contacter à cette adresse mail : [email protected]

Génération Prépa is a digital media platform created by students for students with the goal of helping students succeed in the highly demanding French classes préparatoires and better understand the world of engineering Grandes Écoles.

During my internship, I worked on the development of a deep-learning-based image captioning system designed to assist visually impaired Chinese users by automatically describing visual content. The project combined computer vision and natural language processing techniques to generate meaningful descriptions from images. Key contributions: - Developed a baseline image captioning model combining CNN feature extraction (MobileNetV3) with transformer-based language generation. - Conducted experiments and model evaluation using BLEU metrics to optimize caption quality. - Built an interactive demonstration application using Streamlit to showcase the model’s capabilities. - Presented the project at the Smart City Expo in Taipei, highlighting its potential real-world impact for accessibility technologies.

My work focused on applying machine learning, computer vision, and data analysis to automate complex analysis tasks traditionally performed manually by engineers. Key contributions: - Developed a computer vision pipeline using YOLOv5 to detect key objects during pit stops (tyres, jacks, mechanics) and automatically identify critical time steps in pit stop sequences. - Built a machine learning system to categorize large sets of racing images by event, team, and car component, achieving ~93% classification accuracy and significantly accelerating competitor analysis workflows. - Designed a Python-based user interface to allow engineers to easily explore and analyze image datasets. - Conducted analysis on competitors’ aerodynamic development strategies using CFD data and race imagery, providing insights to the AeroDevelopment team during race weekends.

Worked within the energy performance and analytics team to support data-driven decision making for energy management projects. My role focused on transforming operational data into actionable insights for monitoring and improving energy efficiency. Key contributions: - Designed and developed Power BI dashboards to visualize and analyze energy consumption data, enabling teams to track performance indicators and contractual objectives. - Built a monitoring tool for energy performance metrics used at client headquarters in Paris to support reporting and operational decision-making. - Applied polynomial regression models to forecast weather-related parameters influencing energy performance, improving projections and analysis.

Worked on natural language processing and machine learning systems aimed at improving the interaction between users and enterprise back-end systems. The objective was to transform free-form user commands into structured queries that could be executed across different internal applications. Key contributions: - Developed NLP models (LSTMs and Transformers) to convert user text commands into standardized queries for systems such as CRM tools and SQL databases. - Designed a pipeline capable of handling fuzzy or ambiguous inputs, improving robustness when interpreting user instructions. - Evaluated model performance using BLEU scores and conducted experiments to compare sequence-to-sequence architectures. - Built a text-controlled chess application as a prototype to demonstrate the reliability of the command-to-action translation system.

One month summer internship : I promoted industry during a tour of France.