Hi, my name is
Ghaith Saidani.
I build intelligent data solutions.
Computer Science Engineering student specializing in Data Science and pursuing a dual degree in MS2D. I am passionate about **AI, Machine Learning, and MLOps**, seeking a final-year internship to apply deep technical skills.
View My ProjectsAbout Me
I am a highly motivated Computer Science Engineering student based in **Limoges, France**, with a strong focus on Data Science and Digital Solutions Management. My technical journey has led me to specialize in building robust, end-to-end AI applications.
During my internship at **Capgemini Engineering**, I developed an innovative information extraction system using OCR, OpenCV, and cutting-edge NLP Transformers like **LayoutLM and LLM Qwen**. I am skilled in transforming complex data problems into functional, scalable prototypes using Python and Streamlit.
My expertise spans from low-level performance optimization (like my work migrating Java/JavaFX to QML/C++/Qt at **STMicroelectronics**) to complex deep learning implementations (VAE, Diffusion Models). I am actively looking for a final-year internship opportunity.
Get in TouchTechnical Expertise
Core Languages
- Python (Pandas, NumPy, Scikit-learn)
- C / C++
- Java
- R, SQL
- JavaScript, QML/Qt
AI, ML & NLP
- Keras, PyTorch, Hugging Face
- Generative Models (VAE, Diffusion)
- Transformers (LayoutLM, LLM Qwen)
- Tesseract, OpenCV (Computer Vision)
- librosa, Streamlit
MLOps & Tools
- Docker (Containerization)
- MLflow (Experiment Tracking)
- FastAPI (Model Deployment)
- GitHub Actions (CI/CD)
- Elasticsearch, Kibana, MySQL
Featured Projects
Due Diligence Platform for Crypto Funds
Developed an end-to-end platform for automated crypto fund analysis, integrating advanced ML and NLP models to perform risk assessment and deliver real-time insights. Featured biometric authentication using the MTCNN deep learning model for secure access.
Speech Emotion Recognition System
Implemented and refined generative models (VAE, Diffusion) to enhance emotion classification accuracy on datasets like EmoDB/RAVDESS. The core system uses ResNet classifiers built in PyTorch, with the entire study documented using the CRISP-DM methodology.
MLOps Pipeline with MLflow & Docker
Designed and automated a complete MLOps workflow. This involved containerizing models with Docker, deploying them as REST APIs via FastAPI, and setting up automated CI/CD pipelines using GitHub Actions. Experiments and metrics were tracked seamlessly using MLflow and visualized in Kibana.
Get In Touch
I am currently seeking a final-year internship in Data Science or AI. If you have an exciting project or opportunity, please reach out!
Say Hello