I am an AI Engineer with an M.Sc. in Artificial Intelligence and a background in Business Computing. I specialize in designing and shipping end-to-end AI systems—bridging the gap between raw data ingestion, model research, and production-grade deployment.

My focus is on building models that are not just accurate, but reliable, observable, and maintainable at scale.

• 🌍 Currently based in France (Open to Remote).
• 🔍 Actively seeking CDI opportunities in AI Engineering, Data Engineering, and Data Science.
• ✍️ I write about applied machine learning and MLOps on Medium.

Automated extraction of structured entities (names, dates, expenses) from unstructured administrative documents.

• Fine-tuned CamemBERT/RoBERTa models for French administrative text.
• Achieved high F1-scores by handling low-resource data through custom augmentation.
• Integrated the model into a batch processing pipeline for document ingestion.
• Keywords: NLP Transformers Token Classification PyTorch

End-to-end intelligent OCR system that transforms PDFs into structured financial data in seconds.

• Built a hybrid pipeline using LayoutLM for document parsing and field extraction.
• Developed a FastAPI backend and React frontend for human-in-the-loop verification.
• Fully containerized with Docker for scalable production deployment.
• Keywords: OCR Full-Stack AI FastAPI Computer Vision

Comparative study of agent navigation efficiency in the OpenAI Gym environment.

• Evaluated Tabular Q-Learning vs. Deep Q-Networks (DQN) performance.
• Focused on hyperparameter tuning (epsilon-decay, discount factors) to optimize reward curves.
• Keywords: Reinforcement Learning Deep Q-Learning OpenAI Gym

Binary CNN classifier for chest X-ray diagnosis with a focus on model interpretability.

• Leveraged Transfer Learning (ResNet/EfficientNet) to mitigate class imbalance.
• Implemented Grad-CAM visualizations to provide clinical explainability (XAI).
• Keywords: CNN Transfer Learning Explainable AI (XAI) Keras

• MLOps: Implementing automated retraining loops and experiment tracking with MLflow.
• Vector Databases: Deepening expertise in RAG (Retrieval Augmented Generation) using Pinecone/Chroma.
• Large Language Models: Fine-tuning open-source models (Llama/Mistral) for specialized tasks.

I'm open to discussing AI systems, data architecture, or potential CDI opportunities.

"The best way to predict the future is to build it."