Carrying out data projects to improve performance, solve problems and/or support the challenges of business teams in different countries, according to the priorities identified: receiving requirements, analysing data, creating models and tools. Development of CLI/LIB/API, putting in place MLOps-standard pipelines for CI/CD (GitOps/ArgoCD), model monitoring (MLflow), code quality standards (black/flake8/mypy/pylint/isort) and tests. Constantly performing scientific watch for state-of-the-art proposals, as well as for the maintenance and update of models in production.

I worked on the analysis of users' Quality of Experience (QoE) on the Web, one of the most successful Internet application. Specifically, I had the chance to deal with data-driven models of user satisfaction, built through different measurement campaigns, particularly to assess user satisfaction on Wikipedia pages, in the context of a collaboration with the Wikimedia Foundation.

I worked at the Web Intelligence Group on the measurement of societal bias embedded in transformer-based language models (BERT, RoBERTa, GPT2, etc.) to ensure fairness in conversational systems.

I have worked in the context of the NewNet@Paris Cisco's chair on topics related to Internet measurements, by employing state-of-the-art algorithms of Machine Learning.

The topic of the internship was merging the world of network measurements with that of Machine Learning. The data gatherer module uses the main networking Python libraries, whilst the data analysis one exploits Pandas.

The internship has focused on the development of an algorithm for the geometric verification of images. Given two clouds of data points, each one referring to an image, the algorithm is able to detect whether they are correlated or not, i.e. if the two images depict the same object. This test, executed on pairs of images with ground truth, can provide a score of the accuracy of the visual recognition technique. The work has been carried out in MATLAB.

Thesis: Longitudinal, large-scale and unbiased Internet measurements: the users, the Web, the models.

.

Thesis: Design and experimentation of host classification algorithms using active measurements in the Internet.

Specialist Subjects: Programming for IoT Applications, Operational Research, Statistical Signal Processing, Automotive Infosystems, Management and content delivery for Smart Networks, Mobile and sensor networks, ICT in transport systems, ICT in building design, ICT for Geomatics, ICT for Health, Smart Grids, Mobile Application Development.

Joint bachelor program in Information and Communication Technology Engineering developed by Tongji University from China and Politecnico di Milano and Politecnico di Torino from Italy to train the selected talents in a multicultural environment.

Specialist Subjects: Signal Theory, Computer Networks, Networking Laboratory, Advanced Programming.

Git, GitlabCI, SVN, Apache Hive & Microsoft Office

Python (numpy, pandas, sklearn, transformers, sentence-transformers, pytorch, keras, nltk), Bash, LaTeX & SQL

Italian (native), English (advanced), French (advanced) & Chinese (beginner)

Communication, Team work, Empathy & Adaptability

Flexibility & Creativity

Winner (in team with Andrian Putina and Matteo Romiti) of the Hackaton proposed by 2018 the ACM TMA’18 Conference. The goal was the development of an algorithm for fingerprinting the usage of Internet services, evaluating the implications for users’ privacy.

Winner (in team with Dimitros Milloris, Marcel Enguehard, Andrian Putina and Thomas Green) of the Hackaton proposed by the Rescom’18 Summer School. The goal was the temporal and spatial analysis of the tweets posted mduring the world cup 2014.

Design and implementation of a coherent data analysis framework capable to collect and integrate heterogeneous geo-referenced data in order to investigate the urban mobility. The city of Turin has been taken as case study and for it a remarkable focus has been given to Enjoy and Car2Go car sharing providers. UMAP has been fully developed in Python, mainly exploiting the Pandas library and leaning on MongoDB.

Design, development and emulation of the modules for a Building Management System (BMS) such as shading management, lighting management and air quality control system referred to a real building project carried out by a team of architects. The modules have been implemented in Python.

Design and development of a physical system to monitor both the lighting and the heating in a smart home environment. A Raspberry Pi has been involved acting as controller and as a broker used by a set of Arduinos connected to the sensors. A graphical interface to make the user interact with the system has been implemented using Bootstrap, PHP and Javascript.

Cinema, Theatre, Photography, Literature, Travels, Art, Psychology.