Freelancer profile translated to English.
Back to original languageAbout Mathieu
AI Systems Engineer specialized in the design and deployment of systems leveraging LLMs and Generative AI (GenAI). My work involves transforming language model capabilities into truly usable production applications and systems, with a focus on architecture, robustness, and maintainability. I design RAG (Retrieval-Augmented Generation) architectures, document pipelines, and LLM workflows for efficient use of internal knowledge bases. My positioning is that of a systems-oriented AI / GenAI engineer: connecting models, data, and applications to build complete solutions. Areas of expertise: • RAG architectures and document search systems (vector search, BM25, reranking) • LLM workflow and agent orchestration (LangChain, LangGraph) • Integration of LLMs into business applications • Design of document pipelines and knowledge access systems • Industrialization and deployment of AI systems • AI architecture in constrained environments (GDPR, sensitive data) Recent examples: • Design of an internal RAG + LLM document assistant used by operational teams • Reduction of approximately 30% in queries to subject matter experts through automated knowledge access • Design of a secure document pipeline for using cloud LLMs in a GDPR environment In parallel, I also develop projects around GenAI architectures, notably a GDPR assistant based on a hybrid RAG architecture (vector search + BM25 + reranking) orchestrated via LangGraph. PhD in Physics — analytical and systemic approach applied to AI systems engineering.
French
Native or bilingual
English
Fluent
Can work on-site
Limoges (up to 50km)
Experience
- Softeam,Senior Data ScientistSOFTWARE PUBLISHINGDecember 2020 - Today (5 years and 6 months)Bordeaux, FranceLong-term assignment within the La Poste Group in a highly regulated environment involving large-scale sensitive data processing. AI systems architecture and document pipelines. Industrialization of OCR/ICR pipelines with strict quality, traceability, and auditability requirements. Refactoring of a critical document processing service (C# → Python / OpenCV migration) to reduce technical debt and improve maintainability. Standardization and optimization of technical workflows leading to ~40% reduction in maintenance time and improved overall robustness. GenAI systems development and LLM integration. Design and development of an internal LLM-based document assistant for operational teams. Complete RAG architecture: ingestion, indexing, retrieval, and generation. Azure OpenAI (GPT) integration with FAISS vector database. Development with Python, LangChain, LangGraph, and Streamlit. Collaborative interface allowing knowledge base updates by ~50 users. User feedback system to improve response relevance. Impact: • reduction of approximately 30% in queries to subject matter experts • improved onboarding of new agents. Secure document pipeline for cloud LLM use. Design of a document sorting and anonymization pipeline enabling the use of cloud LLMs within a GDPR framework: • multi-document mail processing • progressive filtering based on business metadata • automated anonymization before exposure to LLMs • automated document routing to business services. AI systems governance and compliance. Design of an anonymization engine ensuring GDPR compliance for processes using external AIs: • risk analysis • re-identification prevention • data flow security • privacy-by-design approach
- NOVAE,R&D EngineerRESEARCHApril 2017 - June 2019 (2 years and 2 months)Aixe-sur-Vienne, France• Simulation of nonlinear equations (MatLab, Python): feasibility studies of optical fiber lasers for industrial and biomedical applications. • Automation of test benches and measurements (Python, LabView): design of automated benches ensuring quality control and reproducibility. Design of fiber lasers (SolidWorks): complete design and assembly of laser sources. Simulation (MatLab – Python) feasibility study of optical fiber lasers for ophthalmic applications. • Automation of measurement benches spatially and spectrally resolved benches, processing of 6+ million points. conferences
Recommendations
Be the first to recommend Mathieu
Help this freelancer shine by sharing your experience working together.
These freelancer profiles also match your criteria
Agatha Frydrych
Backend Java Software Engineer
4.7
(3)
2
Baptiste Duhen
Fullstack developer
4.6
(4)
5
Amed Hamou
Senior Lead Developer
4
(2)
7
Audrey Champion
Web developer
4.3
(3)
4
Education
- MasterCentraleSupélec2020Project driven diploma : - Food quality notation using OpenFoodFacts database to pave the way for a mobile scanning app, - Foretelling energtic consumptions of buildings, - Clustering of a customer database to help targeted marketing, - Automated labeling of questions through NLP process, - Image classification using Deep Learning methods through CNN and transfer learning, - Exploration in Handwriting Text Recognition to process centuries old documents,
- Doctor of Philosophy - PhDUniversity of Limoges2017Bachelor, Master and Ph. D. at Limoges University - Research part at XLIM Laboratory. Doctor's degree in photonics on : "Development of dispersion tailored optical fibers for ultrafast 2 μm lasers" Master's degree obtained with honours and a merit scholarship Bachelor's degree obtained with honours and a merit scholarship