CV

MDPI AG, Basel

• Train, deploy, and evaluate NLP models for academic-writing applications, including domain-specific language model pretraining and classification of unbalanced data.
• Built embedding representations of most academic works indexed by Pubmed, Scopus, and others, powering a scalable recommendation system on a vector database.
• Own on-premise LLM infrastructure at scale: deploying and serving models up to 72B parameters with vLLM and Kubernetes, now serving billions of tokens across internal workflows.
• Use LLMs and RAG to generate sourced descriptions of academic topics, iterating from Llama 2 to Mistral 7B to Mixtral.
• Apply NLP to ethical content checks on submitted manuscripts, and built a vision-language pipeline that generates alt text for academic figures.
• Built a custom RAG system over internal guidelines using Qdrant and Qwen2.5-VL, including custom parsing of source documents.
• Mentored a master's intern on a machine learning project.

MDPI AG, Basel

• Pretrained a BERT language model for applications related to academic writing.
• Built topic and journal recommendation for academic articles, demonstrating the business value of language modeling for the publishing industry.

École Polytechnique Fédérale de Lausanne (EPFL)

• Focus on applied data science, natural language processing, and more.

ENSISA, Mulhouse, France

Antoine de Saint-Exupéry high school

• Focus on mathematics and physics.

EPFL — Human Computer Interaction Group, Lausanne

• Studied how interjections affect users' perception of chatbots.

ENSISA, Mulhouse

• Computer vision project predicting COVID-19 from chest X-rays.