Romain Loury

-Machine learning and deep learning algorithms development to combating fraud with SparkML -Creation of batch scripts to process and integrate large volumes of data with Spark and Hive (until 20 billion transactions) -Dockerized python script development in a sensitive data tokenization project in a PCI DSS environment -Dashboard creation -Creation of a data lake on Hadoop cluster : user law, architecture, additional services ... -Technical reference of Data scientist Team

- Modelled an algorithm that match teachers to students following workshops with the psychologist and the founder of Woash - Studied the different possible cloud architectures as well as the associated costs - Managed a team of 2 people to build the workflow

- Collected, verified and source external databases in order to extend the internal databases to improve the input of predictive bricks - Studied the different possible approaches to make predictions and establish a shortlist with the most suitable solutions - Developed a neural network that meets the objectives and frameworks of the mission and also proposed an architecture for industrialization

-Outsourcing of internal data to a data lake hosted by Azure -Offered different agile architectures in order to efficiently perform the various existing services: processing of daily batches reassembled by all the stores at night and continuous data availability -Automation of the various tasks to be performed in the cloud and ensure reliable storage without loss of data -Opening of Data Science with the creation of a use case on Azure ML and implementation of Twitter streaming with sentiment analysis

- Created relevant clusters grouping the different customer profiles and determined the main characteristics of each cluster - Formed the clusters and generated Sankey diagrams showing the flows between the different clusters over time - Determined the actions to be taken in order to push customers to be more digital and more profitable through an analysis of segmentation