Freelancer profile translated to English.

Description

I design and industrialize AI systems for companies that want to move beyond the demo stage.

My background: eight years of data science in an industrial environment. At Stellantis, I worked on predictive maintenance, predicting failures from over 1.5 petabytes of connected vehicle data. At Safran, as a generative AI referent, I implemented RAG systems and agents capable of querying internal databases in a secure AWS environment. I am currently preparing a doctorate on the detection of defects in power cables using signal processing.

My way of working is roughly as follows: I prefer a slightly less spectacular model whose limitations are known to a beautiful demo that collapses at the first real-world case. Honest baseline, no-data-leak validation, quantified uncertainty, and code that still runs after I leave.

I can help you in three areas:

• Enterprise GenAI: RAG on private data, agents, LLM fine-tuning, clean deployment

• Anomaly detection and predictive maintenance on signals and time series

• MLOps: packaging, CI/CD, drift monitoring

Short assignments (audit, POC) as well as long ones (production deployment). Remote, in French or English. If the need is still unclear, we will frame it together before writing a single line of code.

Languages

French
Native or bilingual
English
Native or bilingual

Workplace preferences

Remote only

Primarily works remotely

SPRK TECH (CIFRE) — GEEPS CentraleSupélec
AI & Signal Processing PhD Student
TECH
March 2026 - Today (5 months)
Gif-sur-Yvette, France
• ● Non-linear defect detection pipeline (chafing) using denoising autoencoder + conformal prediction: 85% recall with a guaranteed False Alarm Rate (FAR) ≤ 1%, outperforming a reference matched filter
• ● Robustness to drift via Adaptive Conformal Inference (ACI)
• ● Online change point detection (BOCPD) with physical prior
• ● Multi-cable domain adaptation
• ● MCTDR (Multi Carrier Time Domain Reflectometry) signal processing: FFT (Fast Fourier Transform) / STFT, wavelets (CWT)
• ● Validation on synthetic data then real industrial data
Machine Learning & Deep Learning Traitement de données Traitement du signal Data Scientist PyTorch Time Series Analysis and Forecasting
SAFRAN
Data Scientist
AVIATION AND AEROSPACE
November 2022 - February 2026 (3 years and 3 months)
Magny-les-Hameaux, France
• ● Intelligent agent (MCP) coupled with RAG, querying heterogeneous databases and the web for contextualized answers
• ● Fine-Tuning on internal data + question-answering on private data via LLM (LLaMA 2, Mistral 7B, GPT-4o, Claude) in a secure AWS environment
• ● Active learning API (NER + classification)
• ● Automatic generation of aircraft assembly procedures (Visual Transformer + NER + LLM)
Machine Learning & Deep Learning Natural Language Processing (NLP) IA générative Time Series Microsoft Power BI
Stellantis
Data Scientist
AUTOMOBILE
September 2018 - September 2022 (4 years)
Carrières-sous-Poissy, France
• ● Failure prediction models (12V and traction batteries) in Big Data context: > 1.5 PB of connected vehicle data
• ● Weak signal detection and Remaining Useful Life (RUL) estimation (LSTM, CNN, Transformers, GMM)
• ● Automated Data Lake ↔ IS pipeline
• ● Machine Learning or ML algorithms (XGBoost, SVM, logistic regression, KNN, MLP) and reconstruction of explanatory variables from telematics data
Machine Learning & Deep Learning Python, Machine learning, Data science, SQL, Data visualisation Time Series Analysis and Forecasting Big Data Microsoft Power BI