As a Consultant within the Accenture Applied Intelligence team my roles are:- interact with client's Data Science and Business teams to estimate feasibility and outcomes of tasks- develop end to end projects- supervise and advice junior resourcesI have worked for multiple Media and Telco clients in ML projects such as: - Churn estimation and prevention- Propensity estimation and prospect clients identfication- Customer Base clustering and and best offering proposal- Resources optimal assignment and other optimization problems

I am working for Telco and Media clients developing machine learning models in order to support business strategies. My expertise ranges from the creation of ad hoc ETL pipelines, development and deployment of machine learning models and visualization of results through Data visualization tools such as Tableau and Power BI.Major International Media enterprise:- Deployment of data pipelines (Apache Airflow) for the automatic production of KPIs at multiple levels of drill down to enhance product performance monitoring.- Development of several live dashboards in Tableau for daily and weekly reporting to management.International multimedia publishing group:- Behavioral segmentation (clustering) of the cookies visiting the website. My task ranged from the creation of the KPI (Impala/Hive technologies), data analysis and KPI validation, modellization (K-means algorithm) and development of production code. The output of the model has used for marketing campaigns.- Development of a classification model for subscriber's churn prevention. In this stream my task was the development of the algorithm (Logistic Regression algorithm) and its validation.Major Italian Telco company:- Mapping of the Customer Base on the Italian territory and related analysis in order to define cross-selling actions.

In June 2015 I was awarded a scholarship from the Ecole Doctorale INTERFACES to study the problem of approximating supply-demand distributions networks under the supervision of professor Antonin Chambolle. Usually the problem is casted in the context of graph theory, where networks are modeled as weighted trees and the optimal network is the one which minimizes a certain cost function. In this framework many instances of the problem are NP-hard, meaning that the time required for finding a solution grows exponentially with the number of locations to be connected. It must be said that the Optimal Control community has developed some fast algorithms that work very well when networks develop only in two dimensions. Nevertheless these methods do not scale well when a third dimension is added, e.g. when dealing with the circulatory system in our body. In my thesis I developed new methods to approach these instances of the problem. Rather than using Optimal Control theory I chose the framework of Calculus of Variations and Geometric Measure Theory in which both the networks and the cost functions are described by means of more complex mathematical objects, a choice which rewards with more flexibility to generalization. More specifically, I have inquired the validity of the approach and conducted numerical simulations on some test problems. This results are contained in two published articles and one currently under review.

Master Thesis on: "A variational approach to the Steiner Minimal Tree Problem" under the direction of Antonin Chambolle.