About.
Hi, I’m Ahmed Samady. I’m from Larache, Morocco, and I work in AI and data science. I did my Bachelor’s and Master’s in AI and Data Science at the Faculty of Sciences and Technologies of Tangier, and currently pursuing a PhD in Mathematics and Decision at the UM6P Vanguard Center. Most of what I build lives at the intersection of machine learning, deep learning, and software engineering. I like working on natural language processing and generative AI, but I’ve also spent a lot of time with agent systems, compliance audits, and real-time data pipelines. Some of my projects include a chatbot for my faculty using Llama 3 and RAG, a Twitter sentiment analysis system built with PySpark and Kafka, and AI agents that check system configurations against security benchmarks. Python is my main tool, but I’m also comfortable with C++, Java, and SQL. I use frameworks like PyTorch, TensorFlow, and LangChain or LangGraph depending on the project. More than anything, I enjoy learning new things quickly and putting them into practice, whether it’s a new model architecture or an entire workflow.
Education.
PhD in Mathematics and Decision
University of Mohammed VI Polytechnic / 2025 - PresentI am currently pursuing a PhD in Mathematics and Decision at the UM6P Vanguard Center.
Master's degree in Artificial Intelligence and Data Science
Faculty of Sciences and Technologies of Tangier / 2023 - 2025I hold a Master's degree focused in Artificial Intelligence and Data Science, which focuses on various fields including machine learning, deep learning, data mining, NLP, big data, and computer vision. This program provides a comprehensive understanding of advanced AI techniques and data analysis methodologies, equipping me with the skills to develop and implement sophisticated algorithms for diverse applications in today's data-driven world.
Bachelor's degree in Computer Science
Faculty of Sciences and Technologies of Tangier / 2020 - 2023I hold a Bachelor's degree in Computer Science, which provided me with a solid foundation in various fields such as data structures, web development, networking, and operating systems. Additionally, I gained a strong understanding of mathematical disciplines, including linear algebra, calculus, statistics, and probability. This comprehensive education has equipped me with the essential skills and knowledge needed to excel in the diverse and dynamic field of computer science.
Experiences.
DataProtect / Casablanca, Morocco
February 2025 - June 2025
InternshipFull-timeOn-siteAt DataProtect I worked on building AI agent workflows and making them practical:
• Put together an end-to-end agentic system that checks remote machines against the CIS Benchmark.
• Built REST APIs around the agent workflow using Flask, waitress, and an nginx proxy.
• Designed a Self-RAG setup with LangGraph and ChromaDB to make the system more reliable.
• Tuned and tested locally deployed LLMs and embeddings with custom evaluation frameworks.
• Created and deployed an internal knowledge base chatbot so employees could query docs and policies with natural language.
PC Halle / Tangier, Morocco
April 2023 - June 2023
InternshipFull-timeOn-siteHere I mixed psychology, game design, and code to improve a mobile educational game:
• Applied reinforcement schedules to the reward system, which cut down player errors and boosted engagement.
• Added performance tracking so we could actually measure learning outcomes and gameplay behavior.
• Built new challenge patterns in Unity and C# that made the game more engaging and improved the flow experience.
Projects.
Content-Based 3D Models Retrieval System
A web application designed to implement a robust Content-Based 3D Models Retrieval system that enables efficient image search and management through visual features. Users can upload, download, delete, and categorize models into predefined classes. The system computes and displays shape descriptors for models, including Zernike moments, and Fourier descriptors, in addition to viewing the model in 3D. It supports a simple search to retrieve visually similar models, providing an intuitive and dynamic way to explore the Pottery dataset.
Content-Based Image Retrieval system based on Bayesian Relevance Feedback
A Content-Based Image Retrieval (CBIR) system that enables efficient image search and management through visual features and relevance feedback mechanisms. Users can upload, download, delete, and categorize images into predefined classes, as well as generate new images by applying transformations like cropping and scaling. The system computes and displays visual descriptors for images, including color histograms, dominant colors, Gabor texture filters, Hu moments, and additional custom descriptors. It supports both basic search to retrieve visually similar images and an advanced Bayesian relevance feedback mechanism to iteratively refine results, providing an intuitive and dynamic way to explore the RSSCN7 dataset, which consists of 2,800 images categorized into seven scene types such as Residential, Forest, and Industry.
FSTT Chat Bot
A chatbot for the Faculty of Sciences and Techniques of Tangier (FSTT) using a combination of retrieval-augmented generation (RAG) and fine-tuning techniques. The chatbot is designed to provide accurate and contextually relevant responses to a wide range of queries related to the academic environment at FSTT. The RAG technique is used to extract information from PDF files and generate responses based on the context derived from these embeddings. The fine-tuning process involves adapting pre-trained language models (Llama 3 8B instruct) to understand and generate text specific to the academic context of FSTT. The chatbot is integrated into a user-friendly interface that allows users to choose between the RAG and fine-tuned models based on their preferences or needs. It was developed using a wide range of tools and technologies, including Hugging Face, PyTorch, Kaggle, Google Colab, Unsloth, Langchain/Langserve, SvelteKit, ChromaDB, MongoDB, and Docker. The architecture of the chatbot consists of three Docker containers: the User Interface (UI) container, the API container, and the Model container. The chatbot is deployed using a MongoDB database to store app-specific data, such as conversations and history.
Twitter Sentiment Analysis
Twitter Sentiment Analysis system that leverages a Kafka and Spark pipeline to ingest and analyze Twitter posts in real-time, providing instant sentiment predictions using a pre-trained logistic regression model with cross validation. The user-friendly web interface, developed with Svelte, allows users to initiate and view sentiment analysis jobs. A Flask-based RESTful API facilitates communication between the interface, the processing system, and a MongoDB database that stores prediction results. The entire system is containerized using Docker for seamless deployment and orchestration with Docker Compose, ensuring high portability and manageability.
ML-Toolkit
The project was designed to address the growing need in the fields of Artificial Intelligence and Data Science for effective tools to analyze and interpret the exponentially increasing amounts of available data. Our application provides a comprehensive platform for data preprocessing, machine learning modeling, and data visualization. Developed using the Python programming language and the CustomTkinter library for the user interface, the application is capable of analyzing datasets and implementing various machine learning algorithms, including logistic regression, decision trees, naive Bayes, support vector machines (SVM), Random Forest, and k-nearest neighbors. This general-purpose toolkit aims to bridge the gap in data analysis capabilities, offering an integrated solution for professionals and researchers.
End-of-studies project: GLUPS-123
The goal of this project was to improve a mobile children's serious game by using different reinforcement schedules to improve the game's reward system. We set out to improve the game to make it more entertaining and instructive for kids ages 6 to 8, taking into account the increasing interest in video game development among students and the availability of online self-learning resources. A lot of changes had to be made, like adding player performance statistics and revamping the reward system. We enhanced the user interfaces and gameplay by using Unity and C# programming, which made the learning process more fulfilling and inspiring.