This repository documents the tasks, notebooks, and presentations I completed during the AI L3 Internship at Orange Digital Center. The internship focused on applying AI techniques to solve real-world problems in Natural Language Processing (NLP), Computer Vision, Generative Models, and MLOps.
This repo includes all my work from the internship, organized into clear sections:
- NLP Notebooks & Models
- Computer Vision (YOLOv8)
- GANs vs VAEs
- MLOps Concepts
- Presentations & Reports
Each folder includes practical implementations, results, and resources that reflect what I learned and applied throughout the program.
- Transformers (BERT): Used pre-trained models for Arabic sentiment classification and explored fine-tuning techniques.
- Text Embeddings: Applied word/sentence embeddings for NLP tasks.
- NLP Preprocessing: Specialized techniques for Arabic, including normalization, token cleaning, and stopword removal.
- YOLOv8: Used for object detection and classification in images.
- GANs & VAEs: Compared two generative models and built example pipelines using PyTorch.
- MLOps: Learned the fundamentals of model deployment, containerization with Docker, and production-level thinking in ML workflows.
Folder: NLP/
arabic-sentiment-using-bert-and-embedding.ipynb: End-to-end notebook for Arabic text classification usingD-Hub_nlp.pdf: Summary presentation of the NLP tasks and findings.
Folder: CV/
CV_Classification_object_Detection.pdf: Project presentation.Modified.ipynb: Object detection experiments using YOLOv8.Kerolos_hani_Presentation.pdf: Summary of my internship experience.
Folder: GANs VS VAEs/
gans-vaes.ipynb: Comparative notebook implementing both models for synthetic data generation.
Folder: MIOps/
MLOps.pdf: Overview of machine learning operations, deployment concepts.Containerization.pdf: Concepts of Docker and container-based workflows.
Through this internship, I worked hands-on with a wide range of AI tools and concepts. I didn't just follow tutorials—I built, tested, and modified real models. I gained confidence in:
- Applying deep learning techniques to Arabic language processing.
- Understanding the architecture and logic behind object detection models.
- Experimenting with generative models and analyzing their performance.
- Thinking beyond notebooks—towards how models are deployed and scaled in real systems.
- Email: keroloshani474@gmail.com
- LinkedIn: https://www.linkedin.com/in/kerolos-hani-data/
- GitHub: https://github.com/keroloshany47