CV

Senior Data Scientist specialising in machine learning models on GCP and AWS. I design and ship end-to-end pipelines across the Telco, Media and Retail sectors.

Experience

Senior Data Scientist · Sky Italia, Milan 2024 — present
  • GenAI documentation migration (GCP): a LangGraph multi-agent pipeline running on GCP that reconciles data-lake documentation split across Atlan and Confluence. A hybrid RAG retriever (BM25 plus ChromaDB embeddings, fused via weighted RRF) surfaces the right business context for each asset, which multiple parallel LLM agents turn into enriched descriptions and column-level docs, validated against a strict Pydantic schema and staged for human review.
  • Next Best Action Engine (GCP): a recommender system built on Learning to Rank algorithms, powering the mobile app homepage. It dynamically sorts mixed-content cards — offers, editorial content, services — in the main carousel to maximise engagement and click-through rates.
  • Stacked upselling propensity model (GCP): a stacking ensemble that predicts customer upselling potential. Multiple prediction windows capture short-term digital triggers and lift conversion rates.
  • Next Best Offer Engine (GCP): a recommender that sorts and personalises daily offers across the entire customer base, optimising ARPU and propensity.
Data Scientist · Everli, Milan 2023 — 2024
  • Subscription churn model (AWS): a churn prediction model for premium service subscriptions, taken to production so the marketing team could run targeted retention campaigns.
  • Stock-out prediction system (AWS): a classification model that forecasts high-risk stock-outs, folding in product-specific and seasonal trends to cut the revenue lost to inventory shortages.
Junior Data Scientist · BIP — xTech, Milan 2021 — 2023
  • Consulted for international clients in the Energy, Telco and Pharma sectors.
  • Customer service LLM analytics (Telco): large language models on AWS analysing recorded customer calls, automating topic and sentiment extraction to improve service quality.
  • Marketing mix modelling (Telco): log-log regression models forecasting annual sales and quantifying the ROI of individual media channels, driving budget allocation strategy.
  • Explainable churn prediction (Telco): a Gradient Boosting pipeline on AWS generating periodic churn probabilities, with SHAP values giving stakeholders actionable root-cause analysis for each at-risk customer.
  • NLP for regulatory compliance (Pharma): unsupervised learning (clustering) applied to artwork labelling data, surfacing common errors and trends to reduce rejection rates from regulatory authorities.

Education

MSc in Computer Science and Engineering · Politecnico di Milano 2018 — 2020
  • Specialisation in Machine Learning and Artificial Intelligence.
  • Final grade: 107/110.
BSc in Computer Engineering · Politecnico di Milano 2015 — 2018

Skills

Machine Learning and Deep Learning NLP and LLMs GenAI, RAG and multi-agent systems Predictive modelling and churn analysis Recommender systems End-to-end ML pipelines

Languages and tools

Python SQL GCP (Vertex AI, BigQuery) AWS (SageMaker, Lambda) Azure OpenAI Airflow Docker Git Jenkins pandas scikit-learn LightGBM TensorFlow Keras LangGraph Pydantic Neo4j ChromaDB Power BI Looker Tableau

Spoken languages

Italian — native English — fluent