Freelancer profile translated to English.

Description

AI Engineer and Data Scientist with over 5 years of experience in AI/ML and 2 years in Generative AI, I support companies and R&D teams in the design, development, and deployment of tailor-made artificial intelligence solutions.

## Main Expertise ##

LLM Fine-tuning, RAG architectures, multi-agent systems, Data Engineering, and ML pipelines.

## Technologies ##

Python, Pytorch, Hugging Face, LangChain, LangGraph, FastAPI, Streamlit, Docker, Qdrant, FAISS, AWS-Sagemaker

## Approach ##

Combining technical rigor with a spirit of innovation to transform complex ideas into reliable, scalable, and production-ready AI applications.

## Types of Projects Completed ##

Development of specialized RAG applications (medical, accounting, insurance, dog training).
Creation of multi-agent systems for complex analysis automation (OCR, FEC, clinical documents).
Implementation of anomaly detection pipelines (insurance fraud, finance).
Applied research: new prediction methods in marine biodiversity (+30% performance vs SOTA).

## My Added Value ##

I master the entire AI value chain – from architecture design to production deployment – always keeping your business challenges, result reliability, and ease of use for your teams in mind.

Languages

French
Native or bilingual
English
Fluent

Workplace preferences

Can work on-site

Montpellier (up to 50km)

Inria
Research Engineer in Deep Learning & LLM
PUBLIC SECTOR
May 2023 - Today (3 years and 1 month)
Montpellier, France
Situation: Development of an innovative approach to predict fish species assemblages using Generative AI (LLMs).

Task: Transform continuous data into an LLM-interpretable format to predict fish assemblages.

Actions:
- Conducted a literature review
- Built a baseline ML pipeline (XGBoost)
- Designed an original dataset
- Fine-tuned LLM models (GPT-2, Llama2)
- Optimized hyperparameters
- Compared performance
- Wrote a scientific paper

Result: Revolutionary predictive method achieving 30% higher performance than traditional approaches. Innovative methodology intrinsically modeling inter-species interactions.

Stack: Python, Hugging Face, PyTorch, Scikit-learn, XGBoost, Dash, Matplotlib, Seaborn, Geopandas, Git/Github
Deep Learning Hugging Face Machine learning LLM NLP
Mutuelle Santé
Data Scientist & Data Engineer for Anomaly Detection
HEALTH AND WELLNESS
January 2023 - August 2024 (1 year and 7 months)
Montpellier, France
Situation: Development of a hybrid anomaly detection system for healthcare reimbursements across a network of 10 mutual insurance companies (2.8M beneficiaries).

Task: Design and implement a complete anomaly detection architecture combining traditional statistical methods and Generative AI (LLMs).

Actions:
- Designed an ETL pipeline
- Developed multi-scale temporal features
- Implemented an ML ensemble approach (Random Forest, XGBoost, SVM)
- Deployed an inference pipeline

Result: High-performance anomaly detection system achieving Recall = 0.40 and Precision = 0.55, outperforming rule-based solutions.

Stack: Python, DuckDB, Polars, Pandas, Scikit-learn, NumPy, Matplotlib, Seaborn, XGBoost, PyTorch, Jupyter, Git/Github, Advanced SQL
SQL Polars XGBoost Anomaly Detection Feature Engineering
Fintech
AI Engineer & Data Scientist
BANKING AND INSURANCE
August 2024 - February 2025 (6 months)
Montpellier, France
Situation: Development of an innovative solution for automated analysis of FEC (Fichier d'Écritures Comptables - Accounting Entries File), combining conversational AI agents and automatic Python code generation.

Task: Design and implement a multi-agent system powered by LLMs to automatically analyze accounting data and generate custom Python code.

Actions:
- Designed a multi-agent architecture
- Built a code generation engine
- Implemented an RAG system
- Created a secure sandbox
- Developed a FastAPI backend with a Streamlit interface

Result: Functional POC reducing analysis time from half a day to a few minutes, enabling non-developers to generate complex analysis code via natural language.

Stack: Python, FastAPI, Streamlit, LangChain, LangGraph, LangSmith, OpenAI API, ChromaDB, FAISS, Hugging Face, Sentence Transformers, Pandas, NumPy, Docker, Git/GitHub, Pydantic
Langchain LangGraph ChromaDB RAG AI Agents