Freelancer profile translated to English.
Back to original languageAbout Arnault
🔍 Data Scientist & AI / Machine Learning Engineer, 10 years of experience
💶 CII Approval: Benefit from a 20% tax credit on my services related to your R&D projects.
🧠 Specialties: NLP, AI, Data Science, Machine Learning, Deep Learning, Large Language Models (LLM / ChatGPT).
🌐 Experience: Edtech, Legaltech, Smart City, Greentech, Fintech, Medtech. Team Lead.
🌟 Focus:
- Social Impact
- Modeling: R&D, PoC, Monitoring
- Management & Strategy
- Data Project Supervision.
💡 Expertise / Services:
- Classification and prediction
- Data Visualization
- OCR / Document Transcription (digital documents)
- Textual Trend Exploration
- Entity or Information Extraction in texts
- Customer Churn Detection
- Customer Segmentation
- Image Analysis
- Anomaly Detection
- Dynamic Pricing
- Time Series Forecasting
- Prompt Engineering
- LLM/ChatGPT Adaptation / Fine-tuning
🎓 Degrees: ENSAE Paris, Paris Saclay.
🌍 Remote work, available for travel. Weekly follow-up.
🌐 Recent clients: WHO, Silvr, World Bank or Kalent AI.
English
Fluent
Spanish
Fluent
French
Native or bilingual
Remote only
Primarily works remotely
Experience
- beta.gouv.frAI Engineer - Legal RAGLEGALSeptember 2025 - Today (9 months)Londres, United KingdomLegal RAG, LLM Agents & Production AI SystemsI participated in the design and deployment of an AI legal assistant platform for the Conseil d'État (Beta.gouv), combining LLM, retrieval augmented, and language model fine-tuning technologies.Key achievements of the Jacepair project:• Hybrid RAG Architecture — Search system combining BM25 sparse and dense embeddings with Reciprocal Rank Fusion, on a corpus of 1.6M+ legal articles and 110K+ court decisions• Multi-source ingestion pipeline — Automated integration of LEGIFRANCE (codes, laws, decrees), ArianeWeb (administrative case law), and ConsiliaWeb (advisory opinions), with versioning and change tracking• Intelligent extraction of legal references — LLM + regex system to automatically identify and structure citations of articles, laws, ordinances, and decisions in legal documents• Multi-LLM provider abstraction — Flexible architecture supporting OpenAI, Mistral AI, and Albert (French sovereign LLM) via LiteLLM, with Pydantic validation for structured outputs• Production-ready infrastructure — Docker Compose deployment with PostgreSQL/pgvector, Qdrant, async FastAPI, Streamlit, and optimized connection management (AsyncPG, pooling up to 50 connections)
- kwarto
On Malt
AI Engineer - Technical Document ExtractionMECHANICAL ENGINEERINGSeptember 2025 - September 2025Londres, United KingdomI carried out a mission as a Data Scientist / ML / NLP Engineer. The project aimed to implement a solution for extracting technical entities from PDF documents related to telecommunication installations.- Extraction of technical information from PDFs- Deployment of open-source LLMs on OVH- Implementation of an automatic extraction evaluation system (via MLFlow)- Development of the fine-tuning system- Development of the annotation/correction platform - LaReserve.techAI Engineer (Volunteering)PUBLIC SECTORMarch 2025 - Today (1 year and 3 months)Londres, United KingdomRAG Fact-Checking, Multi-Source Search & Adversarial EvaluationDescription: Development of production-ready RAG fact-checking systems.I participated in the design and deployment of a comprehensive automated fact-checking platform combining multi-source search, structured generation, relevance evaluation, and iterative red teaming.Key achievements of the Vera project:• Multi-Source RAG Architecture — Dual search pipeline (fact-checking + generic sources) with Google CSE, priority system, automatic fallback, and configurable temporal filtering• Structured Query Generation via LLM — Tool-calling architecture designed to maximize recall, multilingual support• Retriever Metrics System — LLM evaluation of source relevance, baseline comparison, and tracking over time• Source Analysis with Embeddings — Detection of unknown sources (hallucination indicator), measurement of recall and precision• Iterative Red Teaming — Orchestration of adversarial attacks in 4 phases (generation → execution → analysis → improvement), automatic refinement over N iterations• TypeScript/Python Bridge
Reviews
Recommendations
MN
+1
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
- Engineer in Statistics and EconomicsENSAE ParisTech2016Cours (Spécialisation) : • Machine Learning • Apprentissage par agrégation • Gestion de données massives • Programmation • Compressed Sensing • Méthodes séquentielles et chaînes de markov cachées • Méthodes statistiques pour l'économétrie • Econométrie Avancée pour les données qualitatives • Soundings • Modèles statistiques dynamiques avec variables cachées • Bootstrap & statistiques asymptotiques • Statistiques Bayesiennes • Machines Learning Avancée • Data Visualization. Cours (Deux premières années) : • Probability Theory • Advanced Statistics • Non-parametric Statistics • Time Series • Game Theory • Econometrics • Data Analysis • Monte-Carlo methods • Dynamic Optimization • Stochastic Calculus • Macroeconomics
- Master II - Quantitative Economics, specialization in decision theoryUniversité Paris-Saclay2016Cours : • Advances Econometrics for Qualitative Data • Semi Parametric and Non-Parametric Econometrics • Advanced Game Theory • Decision Theory • Monetary Economics • MacroFinance • Information, Transmissions and Communications in Games.