Currently workin on Agentic deployment, Big Data and Cloud
- Insurance Life Policy Simulator
- AI Agent for Revenue Analysis with Groq (LLaMA 70B) on Streamlit
- AI Agent for Financial Analysis Automation with Ollama (Qwen2.5 3B)
- Financial Data Analysis Automation with Python
- LLM-Augmented Workflow with Ollama and LangChain for Revenue Analysis
- Brent Oil VaR using Machine Learning with Python (Master Thesis)
- ResNet Computer Vision Classifier with PyTorch
- Bitcoin Signal Prediction using Random Forest with Python
- Insurance Customer Risk Segmentation with Python
- Insurance Claims Classification & Monte Carlo Simulation with Python
- CO₂ Historical Emissions Review with Python
- Worldwide Life Expectancy Analysis with Python
- Geographical Poverty Analysis in Italy with R
- Automotive Sector Europe — Revenue Analysis
- Energy Production and Coverage in Italy Report
- Designed and deployed LLM-powered agents (Groq LLaMA 70B & Ollama Qwen 2.5) for automated financial and revenue analysis
- Built a Streamlit web app with interactive Plotly dashboards for real-time user-driven analysis
- Implemented full pipeline: data ingestion (Yahoo Finance) → financial metrics (17+ KPIs) → multi-year trend analysis & CAGR → automated reporting
- Enabled scalable analysis across 100+ companies and 10+ sectors via prompt-based interaction
- Engineered a PySpark pipeline on Databricks to process large-scale NOAA weather data from AWS S3
- Implemented data lake architecture (Delta Lake + Parquet) for efficient storage and querying
- Computed multi-level aggregations (continent, country, city) using Spark & Pandas
- Delivered insights through scalable visualizations (Matplotlib, Seaborn)
- Developed Value-at-Risk models on Brent Oil using econometric and ML approaches with backtesting validation
- Demonstrated superior performance of ML models (Boosting) in capturing nonlinear risk dynamics and reducing forecast errors
- Applied Monte Carlo simulations and classification models in insurance use cases
- Built customer risk segmentation (PCA + KMeans) identifying underpriced high-risk clusters to support underwriting decisions
- Trained a ResNet-18 (transfer learning) model in PyTorch for image classification
- Achieved ~98% validation accuracy on a 20k image dataset
- Evaluated model robustness using confusion matrix and ROC curve
- Built an end-to-end financial data pipeline (yFinance → cleaning → EDA → feature engineering → visualization → reporting)
- Reduced analysis time by ~50% through automation and reusable workflows
- Conducted CO₂ emissions and global life expectancy analysis extracting macro-level insights
- Performed geospatial poverty analysis in Italy (R) using mapping, spatial visualization and regression models to uncover territorial disparities