Personal projects

Author: Claudio Paliotta et al. 

Period: 2025

Introduction

This project applies advanced analytical techniques to classify Formula 1 race tracks based on their defining physical and performance-related characteristics. Beyond the sporting context, it serves as an example of how structured thinking, data simplification, and clustering methods can be used to make sense of complex systems. The project reflects a consulting-oriented approach to turning multidimensional data into actionable insights.

Objectives

The project aimed to:

• Provide a clear, structured classification of Formula 1 circuits to reveal underlying patterns in race strategies and performance.

• Demonstrate the value of simplifying complexity through data clustering and feature design.

• Illustrate how methodology-driven insights can support better decision-making in dynamic, high-performance environments.

Methodology

Using publicly available data, key features were engineered to capture the essential character of each circuit—such as speed profile, corner complexity, and flow. A tailored clustering approach was applied, with the optimal number of groups selected based on interpretability and relevance, rather than solely statistical performance.
The result: a segmentation of the racing calendar that reveals recurring design archetypes.

Results

Three distinct circuit types were identified:

• Modern All-Rounders: Versatile tracks demanding adaptability from teams and drivers.

• Classic Speed Circuits: High-speed venues where tire strategy and aerodynamic efficiency dominate.

• Technical Street Challenge: An outlier group represented solely by Monaco, characterized by precision, low margins for error, and minimal overtaking. This segmentation makes it easier to anticipate technical and strategic challenges across the season, serving as a framework to support performance reviews, scenario planning, and competitive benchmarking.

Future Work

Potential next steps include:

• Linking circuit types to team performance and car characteristics.

• Extending the clustering logic to other domains where infrastructure and performance interact.

• Building lightweight tools or dashboards to support comparative analysis and decision-making across similar contexts.

Author: Claudio Paliotta et al. 

Period: 2024

Introduction

The DTPI serves as a quarterly pulse check on digital transformation across the EU, highlighting each country's potential within the digital landscape. Inspired by the annual DESI index, this tool focuses on a streamlined set of indicators—Gross Value Added (GVA) in ICT, labor demand, and employment rate in ICT—providing essential insights into the economic drivers of digitalization.

Objectives

The DTPI’s main goals include:

• Offering a quarterly, composite index that measures digital transformation potential with clarity and timeliness.

• Delivering actionable insights into the digital economy to support high-level decision-making for senior leaders and policymakers.

• Establishing a web app interface for easy accessibility and interactive use.

Methodology

The DTPI compiles quarterly data from Eurostat, with each indicator normalized and weighted equally to produce a focused, composite score. The web app setup, using Streamlit, enables users to explore and visualize DTPI results dynamically. The streamlined structure of the DTPI emphasizes the most relevant economic indicators, making it an effective tool for assessing digital transformation potential.

Results

The DTPI’s web application, powered by Streamlit, transforms complex data into accessible, real-time insights. Designed for intuitive use, the app offers interactive visualizations, allowing users to track and compare each country’s digital transformation potential. Streamlit’s open-source framework simplifies data integration, making this an ideal tool for real-time, data-driven insights.

Future Work

Future directions for the DTPI include:

• Expanding the index to cover a broader set of economic and social indicators.

• Enhancing app interactivity and customization to support deeper analyses.

• Collaborating with industry experts to refine the weighting and adapt the index to emerging digital trends.

Author: Claudio Paliotta

Period: 2013

Introduction

This project explores the development and training of a Convolutional Neural Network (CNN) for classifying handwritten digits. The project is intended as a personal endeavor to gain practical experience with CNNs and machine learning. 

Objectives

The primary objectives of this project were:

•  To build and train a CNN capable of accurately classifying handwritten digits from the MNIST dataset.

•  To investigate the impact of background color on the CNN's performance. 

•  To explore methods for improving the CNN's ability to handle variations in image format.

Results

The project successfully achieved its goals. A two-layer CNN architecture with dropout regularization was implemented and achieved an accuracy of 96.9% for classifying digits with white backgrounds. The project then investigated the limitations of this model by introducing images with black backgrounds, which the model failed to classify accurately. To address this, the training data was augmented to include inverted versions of the MNIST images, resulting in a significant improvement to 97.1% accuracy and enabling the model to handle both black and white backgrounds.

This project highlights the importance of considering diverse training data to ensure a robust CNN model capable of generalizing to unseen scenarios.

Future Work

Further exploration could involve:

• Experimenting with different CNN architectures and hyperparameters.

• Implementing data augmentation techniques beyond simple inversion.

• Investigating the application of this model to real-world handwritten digit recognition tasks.

Author: Claudio Paliotta

Period: 2013

Objective 

Building a small anthropomorphic manipulator for writing.

Description 

The manipulator is composed of:

• Three servo motors HS-425BB (180º rotation);

• One Arduino duemilanove.

• The three servomotors are assembled on a “homemade” plastic structure. In particular, the robot arms have been realized from a plastic panel. The basis is made of aluminum and it has been built by me as well. After the realization of the structure and the motor assembly, the model parameters have been calculated and used for the control design.

A controller based on an inverse kinematics method has been implemented interfacing the Arduino duemilanove board with MatLab/Simulink. The control feedback is applied using positional information provided by the servomotors.

Status

The robot can write with a pencil a few letters of the alphabet.

The anthropomorphic manipulator assembled.

Objectives

Building a small autonomous vehicle for indoor applications.

Description

The vehicle is composed by:

• Four DC motors, one for each wheel;

• Two encoders, one for each front wheel;

• One ultrasound sensor for obstacle detection;

• One IMU sensor;

• One Arduino Mega for controlling the motors and the sensors (the motors are connected to an appropriate motor shield);

• One battery pack for the Arduino Mega powering;

• Raspberry Pi 2 model B.

Status

The vehicle is assembled and functional. At the current stage, it can move forward and steer to avoid obstacles detected by the ultrasound sensor. Basic teleoperations are possible.

Future work

Interface the Arduino Mega with a Raspberry Pi using ROS (Robotic Operative System) to render the vehicle more autonomous. In particular, the Raspberry Pi will be the “brain” of the vehicle. The Raspberry Pi will have the role of planning in real time the task to be executed.