Description

Ingénieur GenAI avec une expérience en production dans des environnements R&D exigeants (Air Liquide, Safran). Je conçois et déploie des systèmes RAG, des agents LLM et des pipelines IA de bout en bout de l'architecture à l'évaluation.

Ce que j'apporte concrètement :

Déploiement d'un assistant RAG en production pour 250+ utilisateurs (−60% de temps de reporting) chez Safran Aerosystems

Système RAG complet chez Air Liquide R&D : LangChain, FAISS/Pinecone, évaluation des hallucinations via RAGAS et LLM-as-a-judge

Pipeline computer vision MedTech à 95% de précision (+24% de gain)

Pipelines CI/CD data (GitHub Actions) : −80% de temps de préparation des données

Stack technique :

LangChain · LangGraph · LlamaIndex · RAG · Multi-Agent Systems · OpenAI API · Mistral · HuggingFace · LoRA/QLoRA · FAISS · Pinecone · RAGAS · LangSmith · Langfuse · PyTorch · Docker · Python

Profils de missions recherchées :

Implémentation RAG, architecture LLM, évaluation de modèles, agents IA, automatisation de workflows IA — en remote ou Paris.

Disponible pour des missions freelance à partir de janvier 2026.

Languages

English
Native or bilingual
Spanish
Conversational
French
Native or bilingual

Workplace preferences

Can work on-site

Paris (up to 50km)

Air Liquide R&D Japan
Machine Learning Engineer R&D
June 2025 - December 2025 (6 months)
Tokyo, Japan
• Designed, built and evaluated a production Retrieval Augmented Generation (RAG) system using LangChain and vector databases (FAISS/Pinecone); measured hallucination rates and validated outputs via LLM-as-a-judge.
• Improved a MedTech facial measurement pipeline from 2D to 3D video, achieving 95% accuracy and 95% user validation (+24% accuracy gain).
• Contributed to a global ML project for molecular property prediction; built scalable web scraping pipelines to collect and enrich datasets for ML and RAG-based models.
• Designed data preprocessing workflows integrated into CI/CD pipelines (GitHub Actions), reducing data preparation time by 80% and accelerating clean dataset production by 70%.
Langchain Retrieval-Augmented Generation (RAG) Python CI/CD LLM
Safran Aerosystems
Software AI Developer
June 2024 - September 2024 (3 months)
Paris, France
• Planned and deployed a RAG-based virtual assistant using Ollama and LangChain to streamline internal knowledge access and reporting workflows.
• Designed predictive analytics models to forecast delivery delays from historical industrial data.
• Delivered AI automation tools saving 500+ hours per week across 250+ active users, reducing reporting time by 60% in a top aerospace environment.
inteligence artificielle RAG Machine learning Python Langchain
Harlocmoi
AI Tech Educator & Content Creator
January 2023 - Today (3 years and 5 months)
Paris, France
• Built a tech-focused content platform translating AI and engineering concepts into educational short-form content; generated 10M+ total views and grew a 100K+ audience.
• Improved engagement by 40% using a data-driven content strategy and continuous optimization.
• Collaborated with tech communities and brands to showcase real-world AI applications in the global tech ecosystem.