Freelancer profile translated to English.

Description

Civil Engineer from École des Ponts et Chaussées, I work as a Lead Data Scientist on innovative projects with large volumes of data, from AI use case ideation to industrialization. Throughout my various projects, I have specialized in NLP, customer-centric approaches, Industry 4.0, and MLOps for the industrialization and maintenance of AI projects.

I am also a trainer at Mines Paris.

For knowledge capitalization and sharing, I regularly write articles on ambitious Artificial Intelligence projects.

Languages

English
Fluent
Spanish
Fluent
French
Native or bilingual
Portuguese
Native or bilingual

Workplace preferences

Can work on-site

Paris 12e Arrondissement (up to 50km)

Ministère des Armées
NLP: Automated matching between a job description and CVs
DEFENSE AND MILITARY
November 2020 - May 2021 (6 months)
Paris, France
Tasks performed:
- Data cleaning and preprocessing
- Data anonymization
- Use of pre-trained algorithms (word2vec, Glove, NER)
- Definition of a WordEmbedding specific to the ministry's vocabulary
- Automation of the entire preprocessing so that the user has a list of the 10 most coherent CVs for a given job description

Results: For a given job description, the algorithm provides the end-user with at least 8 collaborators out of the 10 most coherent collaborators from a pool of approximately 8000 people
Python OVH Machine learning Scikit-learn NLP Jupyter
PSA VENTURES
Predictive maintenance of car batteries
AUTOMOBILE
December 2020 - Today (5 years and 6 months)
Paris, France
Provide a decision support indicator to the user to anticipate a possible breakdown

Tasks performed:
- Definition of operational indicators for short-term maintenance optimization and long-term stock optimization
- Development of an architecture with automated ingestion pipeline, packaging of different algorithms, data storage, and visualization of relevant indicators on PowerBI dashboards
- Data valorization through collaboration between Industry and Data Expert
- Understanding stakeholder needs and adding business meaning to all analyses.

Results:
- Identification of anomalies on time series from sensors
- Highlighting of correlations between battery data and potential breakdowns
- Definition of algorithms defining part wear over time and probability of failure
Apache Spark MLlib Microsoft Power BI Jupyter Scikit-learn Machine learning Apache Parquet
DGFIP
Automated detection of swimming pools and buildings
PUBLIC SECTOR
December 2020 - Today (5 years and 6 months)
Paris, France
Tasks performed:
- Data preprocessing and cleaning
- Definition of AI algorithms for detecting swimming pools and buildings from satellite images
- Implementation of MLOps best practices (feedback loop, monitoring of model and data performance, etc.)
- Scaling of the project and integration of AI analyses into the DGFIP architecture
- Making operational indicators available to the business

Results: Knowledge transfer with the DGFIP's IA Factory, daily time savings for business teams
GitHub Kubernetes Deep Learning Machine learning Google cloud Kibana Elasticsearch