Clément Bazin

Commissioned as a Data Science and Data Engineering expert to design, develop, and integrate a project analysis module into its internal platform, offering managers a predictive and centralized view of their KPIs to anticipate risks of delays and budget overruns.

Expert in Data Science and Data Engineering, with a web development component. Intervention as a cross-functional specialist between data teams and application developers.

Project details and responsibilities:

• Data Engineering:

◦ Development of data pipelines (SQL, GraphQL) to automate the collection, processing, and updating of project information.

◦ Management and transformation of data to make it usable in real-time within the application.

• Web Development and Visualization:

◦ Design and integration of interactive charts directly into the platform using Vue.js, Node.js, and Java.

◦ Development of custom visualization tools for KPIs (performance, costs, progress, risks).

• Development of algorithms and models:

◦ Design of prediction modules (delays, cost overruns) using machine learning and clustering models.

◦ Development of Python scripts for analysis, automation, and direct integration into the web backend.

◦ Deployment of models.

• Project conducted using agile methodology in a team of 10 people:

◦ Estimation of development costs and time.

◦ Hybrid work with web developers.

Impact:

• Implementation of a solution integrated directly into Airbus's internal tools.

• Reduction in the time required for project analysis through centralized and interactive visualization.

• Better anticipation of project risks through the developed predictive modules.

As part of engine optimization at Stellantis, the project aimed to reduce polluting emissions while improving engine performance. Integrated into an Agile team, I contributed to the implementation of automated data pipelines, statistical analysis, the development of Machine Learning models, and the creation of interactive dashboards for KPI monitoring and measurement machine control, in order to industrialize the analysis and exploitation of engine data.

Dual role of Data Scientist (statistical modeling, machine learning) and Data Engineer (pipeline implementation, automation, cloud integration):

Project details and responsibilities:

◦ Centralization of heterogeneous data: Excel files, data from measurement benches, data from sites and internal databases.

◦ Implementation of a Snowflake Data Warehouse infrastructure for data storage and exploitation.

◦ Development of automated ELT pipelines with Airflow

◦ Creation of interactive dashboards in Power BI

Statistical and machine learning analyses:

◦ Advanced preprocessing: correlation methods, ANOVA for selecting relevant variables.

◦ Development and comparison of several models:

▪ Linear (quadratic, cubic, polynomial) and non-linear regressions.

▪ Random Forest, XGBoost, neural networks, kriging methods.

▪ Optimization using gridding and cross-validation to obtain the most robust model.

◦ Construction of predictive tools to model the relationship between engine parameters, consumption, and emissions.

• Multi-objective optimization:

◦ Adjustment of engine input parameters using fmincon, gradient descent, and multi-objective gradient descent.

◦ Analysis of results with Pareto representations

◦ Mentoring a team in an Agile (Scrum) context.

As part of the study of durability and fatigue of mechanical components at Stellantis, I participated in the development of interpolation and statistical analysis algorithms to reposition and validate data from finite element simulations (Abaqus), in order to provide reliable and usable results for optimizing component design.

Engineer in numerical analysis, specialized in the development of mathematical and statistical tools for processing and exploiting large datasets from finite element simulations.

Responsibilities and solutions implemented:

• Processing of large volumes of data from Abaqus calculations (stresses, strains, etc.), where calculation points did not correspond to the required locations for fatigue analysis.

• Development of numerical interpolation algorithms to reposition calculated values at the desired points.

• Application of statistical methods to correct and validate interpolations, thus ensuring the reliability of the results.

• Automation of processing using Python, to efficiently manage massive datasets and make the process reproducible.

Impact:

• Provision of a robust and reproducible methodology allowing mechanical engineers to fully exploit simulation results.

• Significant improvement in the accuracy, relevance, and quality of data used for mechanical component fatigue analysis.