Herve Fabrice Ouandji

Fine-tuned computer vision models and built high-quality datasets through image annotation (CVAT). Worked on the conception of multi-agent systems in transport with MCP.

Designed and implemented a rule-based Kinyarwanda text-normalisation system (Weighted Finite-State Transducers) within the NVIDIA NeMo framework. Covered nine semiotic classes with full noun-class morphological agreement, a robust, production-ready NLP pipeline.

Proposed an AI-based PDF segmentation system using object detection to automate digital archiving.

Built a machine-learning fraud-detection system (logistic regression, RandomForest, XGBoost) with data balancing to improve prediction reliability. Applied SHAP for model explainability, supporting decision-making with the business teams.

Implemented secure back-end systems using Laravel and API integration techniques.

Developed scalable REST APIs with FastAPI and PostgreSQL for data-driven applications.

Pursuing a Master's degree in Data Science, specializing in machine learning, statistical modelling, NLP, and computer vision.

Completed a Bachelor of Engineering in Data Science, focusing on data engineering and machine learning.

Earned a Bachelor of Science in Computer Science, building a strong foundation in computing and software development.

Completed GCE A Level studies, preparing for higher education in science and technology.