Projects
R & Shiny
World Athletics Ranking Dashboard
This application provides a ranking of countries, considering the top 100 results by event, year, and
gender.
Explore the map by hovering over and clicking on countries to discover detailed insights about their top
athletic performances!
R
Race visualizer
A visualization tool designed to enhance the broadcast experience of middle and long-distance races, inspired by the dynamic visualizations showcased by World Athletics in recent championships, particularly for the jump and throw events. The tool provides real-time insights into athletes' pace, split times, and estimated finishing times, offering a more engaging and informative viewing experience. Ideal for broadcasters, this tool aims to bring the excitement of real-time data analysis to viewers, particularly during record attempts. This demo showcases the visualization of a world record-breaking 1500m run.
More informationR
How fast can humans run?
An implementation of "A Simple Random Walk Model for Predicting Track and Field World Records"
A prediction model is built for the men's 100m world record, and human performance limits are estimated
using Monte Carlo simulations.
Original paper
Presentation
Python
CNN Image Segmentation (bachelor thesis)
The Standard Model of particle physics, while being able to make accurate predictions, has been proved to fail to explain various phenomena, such as astronomical dark matter observations. In this project, a machine learning application is implemented with the goal of studying dark matter candidates. Images from Charge Coupled Devices (CCDs) in different experiments located underground are used to test different deep learning algorithms. A U-Net model is trained with Python's open-source library Keras. The model performs multi-class image segmentation in order to detect dark matter particle signals among background noise.
Original title: Application of deep learning techniques to images collected with Charge Coupled Devices to search for Dark Matter.
More informationPython
Image Classification
Image recognition implementation with Keras. A CNN is built and trained with the CIFAR-10 dataset. Two models are trained: one without data-augmentation (77.25% accuracy) and the other with data-augmentation (78.04% accuracy).
More information
Unity, Python
AI vs. Humanity
A collection of games where agents are trained with Reinforcement Learning.
More informationR
EUR/CHF forecast
The goal of this study is to fit an ARIMA model to forecast the EUR/CHF currency exchange rate. Different models are obtained through two trend removal methods: a linear model, and differencing.
More informationReport
R
Analyzing schizophrenia
A Generalized Linear Mixed Model (GLMM) is fitted in order to study the correlated observations in longitudinal data of patients suffering from schizophrenia. The model parameter estimates reveal that ’male’ patients evolve less favorably than ’female’ patients, while age does not have a significant effect on the evolution of prevalence of thought disorders.
More informationReport
R
Credit applicant classification
The goal of this study is to determine, in an automated manner, whether new applicants present good or bad credit risk. 9 different machine learning models are trained to classify applicants based on their credit rating.
More informationPresentation
Report
R, MATLAB, SQL, Wolfram Mathematica, Tableau, Excel, etc.
Other projects
Small projects that I work on during my free time.
More information