Freelancer profile translated to English.
Back to original languageAbout Nathan
🇬🇧 Fluent in English (C2). Feel free to reach out in either language.
I design, deploy, and industrialize AI systems in production: LLM agents, RAG architectures, classic ML models. From POC to monitoring, with a focus on reliability and team adoption.
What I do
- LLM in production: agents, hybrid RAG (dense + sparse + LLM reranking), MCP servers, adversarial evaluation (LLM-as-judge type) (prompt injection, privilege escalation, data isolation)
- Classic ML: multi-class classification on imbalanced data (XGBoost), explainability (SHAP), FastAPI / Kubernetes deployment
- Data & MLOps on AWS: ELT pipelines (Airbyte, Snowflake, dbt), bronze/silver/gold lakehouse, IaC (Terraform, CloudFormation), managed services (S3, ECS, Lambda, RDS), monitoring
- AI Governance & enablement: GenAI usage policies, tool selection, workshops and cross-functional support for Product, R&D, and Customer Success teams
Some recent achievements
- Led the AI committee for a SaaS publisher: defined enterprise-wide GenAI policy, adversarial framework to evaluate agent robustness
- Multilingual recommendation engine (hybrid RAG embeddings + TF-IDF + ANN + LLM reranking) deployed in production
- Automatic classification of bank transactions across thousands of imbalanced classes, reducing manual categorization by 70%
Main Stack
Python, LangChain, OpenAI / Anthropic, MCP, XGBoost, PyTorch, FastAPI, Docker, Kubernetes, Terraform, Snowflake, dbt, AWS, Azure
Profile
Mixed background of research (Master's in Particle Physics, ex-CNRS) and engineering (Data Science CentraleSupélec). Comfortable with statistical rigor as well as industrialization and governance challenges.
Available remotely, occasional travel possible. Response within 1 hour on average.
French
Native or bilingual
English
Native or bilingual
German
Conversational
Can work on-site
Strasbourg (up to 50km), Paris (up to 75km), Basel (up to 50km), Luxembourg (up to 50km), Francfort-sur-le-Main (up to 50km)
Experience
- MyUnisoftLead AI EngineerSOFTWARE PUBLISHINGSeptember 2025 - May 2026 (8 months)Paris, France
- Design and deployment of an XGBoost multi-class classification model (thousands of imbalanced classes) for predicting accounting codes from bank transactions: 70% reduction in manual categorization.
- Development of secure MCP servers and design of an LLM-as-judge adversarial test framework evaluating agent robustness against prompt injections, privilege escalations, and data isolation attacks.
- Led the AI committee: defined enterprise-wide generative AI usage policies, tool selection, implementation of best practices.
- Bi-monthly AI workshop facilitation and cross-functional support (Product, R&D, Customer Success): technical feasibility assessments, practice adoption.
- MajelanData Scientist / ML EngineerENTERTAINMENT AND LEISUREApril 2023 - August 2025 (2 years and 4 months)Paris, France
- Design and development of a multilingual podcast recommendation engine combining a hybrid RAG architecture (Azure OpenAI, TF-IDF, ANN search) and LLM reranking, exposed via FastAPI and deployed on AWS (EKS, Terraform, CloudFormation).
- Setup of an ELT pipeline on AWS (Airbyte → Snowflake → dbt) consolidating millions of records into a bronze/silver/gold lakehouse; creation of analytical dashboards (Streamlit, Metabase) on datamarts for internal teams and users.
- Took over a large, undocumented legacy codebase, supporting critical production services after the development team's departure: reverse engineering, stabilization, and maintenance.
- CNRSResearch InternRESEARCHMarch 2021 - June 2021 (3 months)Strasbourg, France
- Processing, structuring, and visualization of massive datasets from simulations for the Future Circular Collider (FCC) project.
- Exploration of Big Data tools and workflows for scientific analysis.
- Presentation at the FCC Jamboree organized by CERN.
Recommendations
Be the first to recommend Nathan
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
- Master's degree (RNCP Level 7) in Data ScienceCentraleSupélec2022Cursus orienté projets multi-secteurs (banque, distribution, agronomie, santé) : - Nettoyage et préparation de données - Analyse exploratoire (univariée, multivariée) - Réduction de dimensionnalité - Entraînement de modèles supervisés et non supervisés, optimisation d'hyperparamètres, évaluation de performance - Deep learning sur données textuelles et visuelles - Déploiement d'API et de dashboards - Versioning Git - Déploiement Big Data et calcul distribué (AWS) Stack : Numpy, Pandas, Scikit-learn, Matplotlib, Seaborn, Plotly, SHAP, TensorFlow, Keras, LightGBM, XGBoost, FastAPI, Streamlit, Gunicorn, PySpark
- Master's in Elementary Particle PhysicsUniversity of Strasbourg2021Mention Bien. Accent sur les concepts statistiques et probabilistes.