Analytical dashboard focused on inferential statistics. The application performs hypothesis tests (ANOVA and T-test) and real-time correlation analyses to determine the statistical significance of factors such as sleep, frequency, and study hours on students’ final performance.

Tools used:
R, Shiny, ggplot2, dplyr, and tidyverse.
Interactive dashboards with Shiny.
Exploratory Data Analysis (EDA)

https://www.twine.net/signin

Development of an interactive web application for exploratory analysis of nutritional data from various cereal brands. The dashboard enables comparison of key nutritional metrics such as calories, sugars, sodium, and protein across manufacturers.

Tools used:
R, Shiny, ggplot2, dplyr, and tidyverse.
Interactive dashboards with Shiny.
Exploratory Data Analysis (EDA).

https://www.twine.net/signin

Application of Generative AI to unstructured MRI and CT scan reports to build a patient journey dashboard for spine orthopedics within Digital Health. The solution enabled strategic decision-making for Primary Care and Digital Emergency services, resulting in R$3.9 million in additional revenue through optimized resource allocation and an 11% improvement in patient retention.

Tools used:
Python, Vertex AI, Gemini API, Scikit-learn, Pandas, NumPy, Multiprocessing, JSON, and Seaborn.
Google Colab and Google Cloud Platform.
Prompt Engineering.
Looker Studio.
GitLab.
Airflow.

This project aimed to develop a Machine Learning model using Convolutional Neural Networks (CNNs) to classify chest radiography images and predict the Agatston coronary calcium score. The internally developed model achieved a 32% cost reduction and reached a ROC-AUC of 0.75 using the VGG-19 architecture, demonstrating its ability to distinguish patients with different levels of coronary calcification—an important indicator of cardiovascular risk.

Tools used:
Python, TensorFlow, Scikit-learn, VGG-19, Pandas, NumPy, and Seaborn.
Jupyter Notebooks.
Google Cloud Storage and Google Compute Engine.
GitLab.

Development of a sales forecasting model for the Rossmann pharmaceutical retail chain using time series regression algorithms, aimed at cost reduction. The model reduced the average forecasting error from 36% to 4.65% (-31%) for the next six weeks, representing approximately €19 million in additional monthly revenue, using Kaggle data.

Tools used:
Python, Pandas, NumPy, Seaborn, Scikit-learn, SciPy, and Boruta.
Anaconda and Jupyter Notebooks.
Machine Learning with XGBoost Regressor.
Render Cloud.
Flask API.
Telegram Bot API.

https://www.twine.net/signin

Development of a customer segmentation model for a banking credit card portfolio using K-Means clustering, aimed at increasing profitability through targeted CRM strategies. The project grouped 8,950 clients into 8 behavioral clusters based on spending patterns, credit usage, cash advance behavior, and payment profile. As a result, the segmentation enables actionable strategies for revenue growth (upsell/cross-sell), cost reduction (more efficient campaigns), and risk mitigation (loss prevention). The expected annual ROI ranges from ~$30K (conservative) to ~$250K+ (optimistic), with a moderate expected impact around ~$100K/year, subject to A/B testing validation.

Tools used:
Python, Pandas, NumPy, and Scikit-learn.
Jupyter Notebooks.
Machine Learning with K-Means Clustering.
PCA for visualization and cluster interpretability.
Cluster evaluation metrics (Silhouette Score, Davies-Bouldin, Calinski-Harabasz)

https://www.twine.net/signin