Elian Mangin

During my final-year internship at EDF, I designed and industrialized a large-scale RAG system to support the drafting of hazard studies. This included building an end-to-end evaluation pipeline for various parsing and chunking solutions (Docling, MinerU, VLMs, similarity-based chunking, etc.) as well as benchmarking multiple retrieval strategies. I focused on production-grade code quality, automated testing (pytest, Ragas), and pipeline optimization using Streamlit, LangChain, and several API integrations. Through this work, I strengthened my expertise in MLOps and Deep Learning applied to large-scale language understanding.

During my gap-year internship at SNCF, I worked on natural language processing solutions, including fine-tuning large language models and implementing a RAG-based question-answering system. I designed a full FAQ-generation pipeline, which I containerized with Docker and integrated into the internal environment. This project strengthened my skills in model deployment and performance optimization.