Freelancer profile translated to English.
Back to original languageAbout Grégoire
I design, develop, and deploy AI agentsthat perform real actions within enterprise systems: ticket creation, report and communication generation, request validation, repository classification.
Since late 2022, I have been building multi-step agentic architectures (prompts, memory, tools, guardrails) connected to internal systems via APIs and webhooks: ITSM, CRM, ERP, SQL databases, data platforms.
Three production references:
- TDF (telecom infrastructure, multi-site): a multi-agent system that drafts crisis communications (90 operators, 24/7) by cross-referencing ServiceNow, Oracle, and SharePoint (RAG). Currently used for real-time crisis communication.
- Sanofi R&D: AI classification of 53,000 laboratory equipment across 7 sites in 4 countries. 342 categories reduced to 22, delivered in 3 months, in autonomous production for one year.
- InsurTech: structured extraction from complex insurance documents by specialized agents, 97% success rate, business users autonomous without technical intervention.
My method:Scoping with business teams, first functional iteration in 1 to 2 weeks, human validation loops (human-in-the-loop), then industrialization with IT teams. Systematic human review and traceability whenever the issue is legal or financial. Data governance integrated from the design phase: access control, anonymization, self-hosted sovereign modes when data is sensitive.
Stack:Python, LangGraph, LangChain, LangSmith, MCP, RAG, SQL, Anthropic, OpenAI, and Mistral APIs (including fine-tuning), Power BI, Streamlit, Azure. Code-first approach for robustness, comfortable with third-party orchestration platforms when the context is suitable.
15 years of experience, including 10 managing a digital agency: I bridge the gap between technical, product, and business teams.
Native French, professional English.
French
Native or bilingual
English
Native or bilingual
Can work on-site
Paris (up to 50km), Orléans (up to 50km)
Experience
- TDFLead AI Engineer · AI Agents ProductionTELECOMMUNICATIONSNovember 2023 - Today (2 years and 9 months)Romainville, FranceTDF operates over 9,200 sites in France (radio, TNT, fiber, 5G) with a network operations center (NOC) running 24/7 with 90 operators and thousands of field interventions per year.Mission: Build the AI team from scratch, design solution architectures, develop them, and then industrialize them with IT teams (ServiceNow, DataMiner). 6 AI solutions delivered to production.
Projects Delivered:
- Automated Incident Communications (Multi-Agent):An agent system connected to ServiceNow, Oracle, and SharePoint (RAG) that generates incident opening, tracking, and closing communications for the NOC. Currently used for real-time crisis communication.- Automatic Site Access Validation:An agent that cross-references 30,000 access requests per year with active prevention plans, makes an reasoned decision, and logs each decision in JSON in ServiceNow. Systematic human review for legally sensitive cases.- Critical Email Processing (ENEDIS stream):Classification of over 4,000 annual electricity outage alerts, structured extraction, automatic creation of ServiceNow tickets. Zero manual operations for standard cases.- Intelligent Alarm Analysis:Predictive model for broadcast infrastructure alarms, reducing false alerts by 25%.- Augmented Data Analysis:Fine-tuning of a Mixtral model generating SQL queries on a schema of 500+ ServiceNow tables, automatic visualization generation, natural language data exploration, performance management Power BI dashboards.Stack:Python, LangGraph, LangChain, MCP, RAG, Anthropic Claude, OpenAI, Mistral (Mixtral fine-tuning), SQL, ServiceNow (ITSM/TSM), REST APIs, webhooks, Power BI, human-in-the-loop, guardrails - SANOFIAI Engineer · AI Classification of a Fleet of 53,000 EquipmentPHARMACEUTICALS INDUSTRYMarch 2025 - October 2025 (7 months)Vitry-sur-Seine, FranceSanofi's R&D scientific equipment fleet (53,024 equipment, 7 sites, 4 countries) was fragmented into 342 categories and 989 sub-categories, lacking a coherent global view. Manual methods had failed.Solution Designed and Developed:Classification using multi-model generative AI (OpenAI, Anthropic Claude), graph matching with confidence scoring, automated quality control, scraping manufacturer websites to enrich records. 3 to 4 human-in-the-loop validation cycles with senior scientific experts.Results:- 53,024 equipment classified across 7 sites in 4 countries- 342 categories reduced to 22 (-93%), 989 sub-categories reduced to 387 (-61%)- Delivered in 3 months (initial estimate: several years), 80-90% accuracy- In autonomous production for one year, over 100 users- Project co-presented publicly by Sanofi at the Adopt AI 2025 trade showInternational context, multi-country coordination in English (France, Germany, USA, Belgium).Stack:Python, OpenAI, Anthropic Claude, graph matching, scraping, ServiceNow, SQL, Power BI
- KUBOTAAI & Data ConsultantRAW MATERIALS INDUSTRYSeptember 2022 - November 2025 (3 years and 2 months)Paris, FranceRole:AI & Data ConsultantSector:Industry / Equipment DistributionContext:Kubota had a catalog of 1.5 million product listings with insufficient raw content for e-commerce: missing or poor descriptions, absent SEO metadata, unexploited images.Role & Responsibilities:Design and deployment of an AI solution for large-scale automated product listing enrichment. Project management from the audit phase to production deployment.
Projects Delivered:
- Automated Product Listing Enrichment— Transformation of 1.5 million raw product listings into optimized marketing content through automatic product image analysis and AI text generation. Each listing enriched with commercial descriptions, technical specifications, and SEO metadata.- AI-Powered SEO Optimization— 35% increase in organic traffic through automatic optimization of over 600,000 product listings (titles, descriptions, alt tags). Results measured on actual e-commerce site traffic.Results:- 1.5 million listings enriched- +35% measured organic traffic- 600,000+ listings SEO optimizedStack:Python, Generative AI (image analysis + text generation), AI SEO
Reviews
Recommendations
JD
FU
+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
- Machine Learning Specialization Certificate, Machine LearningStanford University School of Engineering2022Machine Learning Specialization Certificate, Machine Learning
- Certification Data Analyst, ( SQL / PYTHON / TABLEAU )DataBird2022Certification Data Analyst, ( SQL / PYTHON / TABLEAU )