Stefano Perasso, Ph.D. Email & Phone Number

Paris, Île-De-France, France

Paris Area, France

Paris Area, France

Paris

★ My activities at Data2i:- I worked on the data science projects of Data2i’s clients, entities from various sectors (city councils, e-commerce, industry, logistics, energy, banking) - I was the main contributor to Data2i’s R&D project★ The main data science subjects I’ve been working on:- Anomaly detection in sensor data and logistics- Regression problems: predicting the quantity of garbage produced in one of the most important French cities- Text mining for automatic product classification- Clustering of mixed data types for analyzing a vast list of documents- Operations research for solving logistics problems- Interactive and dynamic data visualization- Creation of dashboards and web applications - Configuration of Solr search engines for several tasks ★ The algorithms I used the most:- Classification: SVM, neural networks, and gradient-boosted trees (xgboost)- Regression: linear regression, random forest and gradient-boosted trees (xgboost)- Custering: k-Means and hierarchical clustering, Gower distance with k-Medoids, self-organizing maps (SOM)- Text Mining and Natural Language Processing: standard text normalization techniques (like stemming, lemmatization, n-grams), One-hot encoding with tf-idf, Latent Semantic Analysis, Non-negative matrix factorization- Dimensionality reduction: PCA, kernel PCA, T-SNE, auto-encoders- Operations research: genetic algorithm- Statistical tests★ The tools I used the most:- Languages: R, Python and SQL- Dashboards and web apps: R Shiny- Data viz: leaflet, plotly, ggplot2, dygraph, DT, D3.js- Search engine: Solr- Streaming platform: Apache Kafka

Paris Area, France

My most relevant accomplishments:★ Work of the DarkSide Monte Carlo simulation program, an object-oriented program written in C++ and based on Geant4 (http://geant4.cern.ch/): - Development of several classes - Code verification and simulation testing on real data - Documentation and user support ★ Modeling of the Liquid Argon (LAr) response to particle energy depositions: - Statistical study (data fitting) of all the published LAr response data - Writing of a C++ program for the simultaneous treatment (fit) of more than 10 different data sets - Implementation of the model in the DarkSide simulation program★ Analysis of the Double Chooz data (several hundreds of TB) to study the detector trigger efficiency. I wrote: - A C++ program for the analysis - Perl scripts for managing the analysis and the job submission to the batch queue★ Test of a photo-detector (Silicon Photo-Multiplier): - Illumination of the Silicon Photo-Multiplier with an LED at room temperature - Statistical analysis of the output and modeling of the noiseAll the analysis programs are based on the Root toolkit, an object-oriented framework for the analysis of big data (http://root.cern.ch/drupal/)I presented talks and posters at the collaboration meetings and at the following conferences:- Neutrino 2014, Boston (MA), USA- GDR Neutrino 2013, Lyon, France- IPRD13, Siena, Italy- Rencontres de Blois, Blois, France- Low Radioactivity Techniques 2013, LNGS, Italy

Greater Philadelphia Area

My most relevant accomplishments at Drexel University:✓ Statistical study of a detector-induced noise in the Double Chooz data and definition of an effective rejection technique. It required the implementation of C++ programs for the data analysis.✓ Optimization of an LBNE detector component (light collector):- Identification of the best light collector shape via simulations- Simulation-based study of the light collector impact on the detector response- Introduction of a more detailed detector geometry in the Monte Carlo simulation program, an object-oriented C++ program based on Geant4 (http://geant4.cern.ch/), and implementation of new classes- Implementation of Perl scripts to manage the simulations and the jobs submission to the batch queue- Implementation of C++ programs for the analysisAll the analyses have been performed using the Root data-analysis toolkit (http://root.cern.ch/drupal/)✓ Validation of the Double Chooz calibration system:- Hardware tests to verify the system operation- Test and validation of the control program (LabView)✓ Participation to the organization of the "Advances in Neutrino Technologies" conference, held in Philadelphia in September I presented talks at collaboration meetings and at the following conferences/workshops:- Advances in Neutrino Technologies 2011, Philadelphia (PA), USA- Project X Physics Study, Fermilab, Batavia (IL), USA

Genoa Area, Italy

My most relevant accomplishments as graduate student:★ Analysis of the Borexino data (big data) to perform several studies:- Optimization of an algorithm for the particle identification- Simulation- and statistical-based rejection of the set of data corresponding to a particular particle type- Data selection and fit★ Development of the Borexino Monte Carlo simulation program, an object-oriented C++ program based on Geant4 (http://geant4.cern.ch/)- Improvement of several methods- Optimization of the optical processes simulation, i.e. tuning of more than 20 parameters on the calibration data★ Calibration of the Borexino detector:- Simulation-based study of the radioactive sources- Participation to the calibration data takingAll the studies were performed on a statistical basis and using the Root toolkit (http://root.cern.ch/drupal/). They required the implementation of C++ programs and Perl scripts.I participated to the data taking operations for the Borexino experiment, also with coordination responsibilities.I presented my work at collaboration meetings.I participated to the following summer schools:- International School on AstroParticle Physics: Dark Matter, Madrid, 2008- International Neutrino Summer School, Fermilab, Chicago (IL), 2009

Laurea Magistrale, Physics, Degree

I graduated in experimental particle physics with a thesis work on the measurement of the radioactive contamination in the detector.

Education record