Freelancer profile translated to English.

Description

Senior Data Scientist specialized in Generative AI, Machine Learning, and MLOps.

Varied experiences: fraud detection, forecasting, NLP, tabular data processing.

End-to-end mastery: modeling (XGBoost, LSTM, Transformers), AI pipelines (RAG, embeddings), industrialization (Docker, FastAPI, CI/CD).

Industry field of expertise

Languages

French
Native or bilingual
English
Fluent

Workplace preferences

Can work on-site

Paris (up to 50km)

BOUYGUES CONSTRUCTION
Data Scientist Consultant
January 2025 - March 2025 (2 months)
- Scoping business needs and developing the data platform roadmap
- Facilitating business needs gathering workshops for various business teams.
- Identifying pain points and formalizing use cases to build a roadmap for a centralized AI data platform.
SAFT TOTAL ENERGIES
Data Scientist Consultant - NLP, Gen AI, LLMOPS
September 2024 - March 2025 (6 months)
Project: Automation of HSE anomaly report processing with LLM and RAG
Development of an intelligent system for automatic analysis of field safety reports, combining the RAG (Retrieval-Augmented Generation) approach with language models (LLM).
Key achievements:
Semantic classification of anomalies (electrical, mechanical, behavioral)
Extraction of critical information (location, severity, corrective action) via structured prompts and LLM
Automatic generation of summaries and recommendations from similar cases (RAG)
Intelligent routing of reports to relevant teams based on typology and criticality
Technical environment:
LangChain, FAISS, OpenAI, Streamlit, PromptTemplate, RetrievalQA
Results:
60% reduction in manual processing time
85% alignment between automatic and human prioritization
Deployment of an interface accessible to HSE teams
LLMOps LLM MLOps Databricks Azure Databricks Python Github Actions Langchain RAG BERT Bertopic OpenAI
AZERION GROUP
Lead Data Scientist - MLOPS - NLP
ENTERTAINMENT AND LEISURE
January 2021 - January 2024 (3 years and 1 month)
Supervision of 3 junior profiles, code review, MLOps and modeling best practices, technical mentoring. Real-Time Advertising Bid Optimization
• Business Objective: Maximize the performance of programmatic real-time bidding (RTB) while improving gross margin and win rate.
• Technical Approach:
• Development of a contextual multi-armed bandit algorithm (optimized exploration/exploitation based on user profile + page/ad context)
• Implementation of a complete MLOps pipeline: experiment tracking with MLflow, data management with DVC, CI/CD with GitHub Actions, deployment via FastAPI & Docker
• Real-time monitoring via Grafana + Prometheus
• Results achieved:
• +18% gross margin
• +15% ad bid win rate
• Technical Stack:
• Python, Pandas, NumPy, MLflow, DVC, GitHub Actions, Docker, FastAPI, Grafana Contextual Ad Targeting Optimization (POC)
• Design of an NLP pipeline focused on contextual advertising: extraction of key themes from web content (articles, product pages) using TF-IDF + semantic embeddings.
• Document vectorization with SentenceTransformers, followed by a matching search between content and ad typology (approximate nearest neighbors).
• Creation of a context-ad relevance scoring based on a supervised model (LogReg / XGBoost).
• User segmentation (KMeans clustering / PCA) to cross-reference profiles & themes.
• Close collaboration with the marketing team for cluster interpretation and targeted activation.
• Technical supervision: coaching, notebook review, A/B test structuring.
• Results
• Increased CTR on several tested segments.
• Better ROI through more contextual and dynamic targeting.
• Strengthened alignment between data and marketing teams.
Nov 2017 - Dec 2020
Reinforcement Learning Awss3 MLOps MLflow Gitlab CI/CD NLP