Experience

PredExIA

• France
• IT Services and IT Consulting
• 1 - 100 Employee

• Senior Data Scientist

  • Jun 2021 - Present

  1. Pour la CMA-CGM (Marseille) Mise en place d'un algorithme de détection de fraudes à partir de différentes sources de données - numérique : codification internationale d'asset, poids... - catégorielle : code containeur, zone géographique, type de conditionnement - textuel : déclaration sur le bon de livraison de l'exportateur (NLP) Principaux algorithmes mis en oeuvre : Random Forest, Isolation Forest, Régression logistic, Boosting (CATBoost) Technologies employées : Python et PySpark Infrastructure : Dataiku Show less



TechMahindra

• Israel
• Information Technology & Services
• 1 - 100 Employee

• Data Scientist

  • Feb 2019 - Jun 2021

  1. Pour Airbus Skywise et dans une philosophie DevOps : - Définition de supports pour les formations des équipes (France et Inde) relatifs : o Au système étudié (ATA36 – Bleed A320), o Aux opérations de maintenance avion (PFR, DAR, LogBook, MEL…), o Aux opérations avions (flight phase notamment). - Développement de modèles de maintenance prédictive o Réalisation de modèles de maintenance prédictive destinés à anticiper les pannes bleed sur les avions et à optimiser les opérations de maintenance o Analyse des données, implémentation des pipelines de transformation des données, optimisation des modèles prédictifs, évaluation de leur qualité o Données utilisées : séries temporelles des capteurs embarqués IoT (DAR), historique des opérations de maintenance, rapports textuels des intervenants (LogBook, PFR, MEL) o Principaux algorithmes utilisés : Random Forest, Gradient Boosting, LSTM, ARIMA Technologies employées : Spark via l’api PySpark en Python ; scikit-learn, XGBoost, Tensorflow v2 pour les modèles LSTM, XGBoost, Hyperopt, stats-model pour les modèles ARIMA, FoundryTS (librairie propriétaire Palantir pour les time series) Infrastructure utilisée : plateforme Palantir (foundry) de Airbus (Skywise) 2. Pour Airbus helicopter à Marignane (mission assurée à distance dans les locaux de TM Colomiers avec possible déplacement sur place) : - Définition/ spécification de vues de reporting sous Spotfire avec le client relatif au suivi d’avancement des inspections en atelier des roulements des satellites de la boite principale des rotors d’hélicoptères (Spotfire, Excel) - Préprocessing des données d’entrées - Déploiement de la solution finale. 3. Aide à la définition de support de présentation d’une nouvelle approche de maintenance prédictive mettant en œuvre les algorithmes de machine learning et à l’identification de clients potentiels (AIL, MRO…) Show less



ALTEN

• France
• IT Services and IT Consulting
• 700 & Above Employee

• Data Scientist

  • Jan 2017 - Feb 2019

  For AiRTHM application (part of the Skywise overall solution based on PALANTIR platform), ensure A/C Health Monitoring of the Airbus fleet defining predictive solutions from in-service data (ACMS data –Reports, SAR files…-, Uplinks, CMS data - Logbooks, CFR, PFR, CMS recall,…). The job consists in:- Following feasibility review, defining solutions (corrective and/or predictive) implementing data mining methods (Isolation forest, Hodrick-Prescott filter, PCA…): o Python prototyping in PALANTIR or stand-alone (Spyder - Python), o Firsts results presentation to DO and SEE specialists for algorithm principle validation, o Algorithm specification for development and integration in AiRTHM, o Algorithm Verification and Validation.- Specifying results display and Look&Feel for final users (Airline MCC, Support department) in PALANTIR and/or AiRTHM.- Establishing recommendations card to fix the failure (TAT) or anticipate its occurrences (validated by Design Office and Customer Support Specialists) and ensuring cards review meeting with OpsCenter (FleetWatchers).- Ensure algorithms follow-up for potential improvement (algorithm coverage or false positive/ false negative improvement). Show less

• R&T @MOST Prognostic & Health Management System Designer

  • Mar 2016 - Dec 2016

  For new aircraft program 2025:- Prognostic design directives definition for Design Office. These directives are derived from different studies like:o Identification of study cases to spotlight prognostic concept for weight savings problematics. Objective is to replace hardware redundancies by prognostic function;o Scheduled maintenance (MPD) study to improve monitoring capabilities (sensors installation, condition monitoring, display/ warning). From systematic preventive maintenance to on-condition/ predictive maintenance.- Define prognostic use case for TRL3 demonstration of IVHM concept applied on aircraft. Study case based on in-service data. Show less



Aeroconseil

• France
• Aviation and Aerospace Component Manufacturing
• 100 - 200 Employee

• R&T @MOST project Wireless Portable Maintenance Access Terminal (PMAT) System designer

  • Jan 2014 - Feb 2016

  In the scope of the @MOST R&T project that aims at defining a new maintenance function for new A320-iD aircraft (SA-LR) and to improve A380 and A350 current solution:- State of the Art of the current existing solution in Airbus and competing solutions study whatever the covered function,- Operational scenarios definition to highlight assess architectures gains (in terms of time savings and costs reduction) of the different architectures,- Define all the alternate architectures taking into account functional and customers (Airlines) needs,- Update/ establish Safety analysis,- Follow and ensure the operability of the envisaged solution,- Drive prototyping development for TRLs needs.During this period, I have participated to answer several tenders and to develop new businesses to help AEROCONSEIL in its diversification needs. Additionally, I had the occasion to ensure project management activities (deliverables follow-up and quality, delivery delay, activities dispatch…).Training avionics systems exploring (5 days). Show less

• R&T @MOST Prognostic & Health Management System Designer

  • Jun 2012 - Dec 2013

  In the scope of the @MOST R&T project that aims at defining a new maintenance function for future aircraft: - Prognostic concept definition applied to aircraft - Board-Ground prognostic architecture and interfaces definition - Data Security Assessment (FIA) - Communication costs analysis - Redaction of a document identifying equipment parameters (and its associated features) to monitor and to prognose it - Integration/deployment at AIRBUS MCC of prognostic algorithms previously developed and tested for TRL needsDuring this period, I have participated to answer several tenders and to develop new businesses to help AEROCONSEIL in its diversification needs. Additionally, I had the occasion to ensure project management activities (deliverables follow-up and quality, delivery delay…).Training English improvement (10 hours by phone. Meeting management, technical presentation, technical conversation with suppliers… Show less

• A380 NSS platform System Designer

  • Jan 2012 - Jun 2012

  Integrated in the A380 NSS platform Team my job consisted in the improvement of the current maintenance scenarios (TSMs) related to the troubleshooting of NSS database and SCI. To do so, the following activities shall have been done: - Define a common TroubleShooting philosophy - Identify among existing TSMs, the redundant ones and those inefficient - Propose a unique TSM to Troubleshoot la Database (idem for SCI) - Consolidate the proposed scenarios with the different Design OfficeDuring this period, I have participated to answer several tenders. Show less

• R&T @MOST Architecture Integration Verification & Validation Engineer

  • Jan 2011 - Dec 2011

  In the scope of the @MOST R&T project that aims at defining a new maintenance function for future aircraft: - Definition of an IVV strategy applicable to the overall project - Definition of the different IVV drivers - Definition of the Batching Policy - Participation to the definition of End to End operational scenarios

• R&T @MOST Systems & Functional Architecture Designer

  • Sep 2010 - Dec 2010

  In the scope of the @MOST R&T project that aims at defining a new maintenance function for future aircraft: - System and Functional architecture definition and modeling - System and Functional interfaces modeling (FID, SID) - Data Security requirements modeling and follow-upof the overall functions contributing to the maintenance of the aircraft (diagnostic, prognostic, configuration,…) on a SysML specification tool (Topcased)



Laboratoire Génie de Production

• PhD Candidate on Industrial Systems

  • Jul 2006 - Aug 2010

  Activities were done in the scope of the Archistic R&T Project leaded by Airbus that aims at improving the availability of future aircraft like A350 XWB. - State of the art of the different architecture dedicated to the control and the monitoring of Large Scale Complex Technical Systems (LSCTSs) - Definition of the problematic and of condition-based maintenance and health status assessment of those systems and its equipments - Analyze of the diagnostic and prognostic typology leading to the assessment of the health status of LSCTSs - Study related to decentralized diagnosis and analyze of the contributions of the ICT and of the distributed technologies for the diagnosis - State of the art of the distributed diagnosis and works relative to this kind of deployments - Study of the limitations of the centralized and decentralized diagnosis approaches and presentation of the added value of the distributed one - Description and formalization of the information and knowledge that support the diagnostic and the prognostic function for their modeling and their exploitation - Proposition of a characterization for the different status of a component - Description of requirements for the monitoring layer of the LSCTSs in order to implement the proposed diagnosis and prognosis principles that are then specified by the means of algorithms - Proposition of an health assessment method for the LSCTSs - Development of a simulation platform to evaluate the performances of the centralized and the distributed deployments - Definition of use case common to the two deployments to compare their performances Title of the dissertation: "Distributed architectures for diagnosis and prognosis of large scale complex technical systems" Show less