Nephtali Garrido-Gonzalez

Created a machine learning model (gradient boosting) to detect risk factors on the most common type of income tax evaders in Mexico. Used and trained a graph neural network to detect networks of companies used for tax evasion, with over 60% accuracy.

A gradient-boosted tree model to predict the total number of days a car will spend in inventory before selling, based on the characteristics of the car and the number of page views it receives within the first few days online. In principle, we could predict inventory days per showroom with a mean error of 29 days, giving us a tool to select the showroom with the shortest inventory days.

A collaboration with the online learning platform Netzun. This introductory course to deep learning, aimed at students and early-career professionals, teaches the basics of the deep learning algorithms and guides the student step by step on creating their first neural network.

Developed an interactive Python application for comprehensive analysis of Doppler ultrasound images. The application provides a user-friendly graphical interface (GUI) that allows medical professionals to load, crop, and analyze Doppler waveforms from ultrasound images of carotid arteries.

Key features include:
• Image loading and cropping.
• Scale calibration.
• Automated waveform analysis:
- Peak systolic velocity (PSV) and end-diastolic velocity (EDV) calculation;
- Acceleration and deceleration time analysis;
- Flow calculation;
- Pulsatility and resistive index calculation.
• Interactive visualization.

The technical stack I used includes Python, OpenCV, Tkinter, Pillow (PIL), NumPy, SciPy, and Matplotlib.