About me

I am a mathematician and programmer focused on data science. I work with large datasets, using statistical techniques to extract useful insights that support strategic decision-making. I have implemented machine learning models for behavior prediction, customer segmentation, and resource optimization in various types of projects. My approach is to deeply understand the client’s problem, translate it into mathematical terms, and develop automated solutions that deliver value, improve efficiency, and produce measurable results.

BankEase Application

BankEase is an application designed to provide users with an intuitive way to explore and analyze their spending habits based on their banking transactions.

Technologies used:

• Intelligent chatbot: An interface with a RAG system connected to an LLM, capable of answering questions about the bank’s documentation. A semantic router enhances decision-making by selecting the most appropriate action for each query.
• SQL retrieval system: automatically generates queries on the transaction database from natural language inputs.
• Fraud detection module: ML-based model that identifies and alerts suspicious transactions.
• Personalized recommendation system: supports financial decision-making based in the history of transactions.
• Manager dashboard: administrators can analyze aggregated consumer behavior across all users.

BankEase combines NLP, information retrieval, machine learning, and data visualization to deliver an integrated experience for both individual users and business managers.

Clinical Patient Management Software

This web-based application is designed to help healthcare professionals maintain detailed clinical records for their patients. It allows doctors to store and retrieve information such as temperature, weight, and treatment details for each medical visit. The system also includes functionality to generate prescriptions in PDF format for easy sharing and printing. The system is connected to a cloud-based database, ensuring that all patient data is securely backed up and accessible at any time. The backend is built with Flask and leverages SQL for efficient database management. Doctors can access their accounts from any web browser using secure login credentials, eliminating the need to install software locally.

Customer Segmentation on Bank Credit Data (K-means & Affinity Propagation)

The objective of this project was to segment customers of a German bank into groups with similar behaviors using clustering techniques. This helps the bank understand its customers better and make more informed decisions about financial products they could offer them.

This type of analysis is crucial because it helps financial institutions better understand their customers and how they interact with their products. Instead of treating all customers the same way, clustering allows for the creation of specific segments that can be worked with more effectively. This not only improves customer relationships but also enhances profitability and operational efficiency for the bank.

Ecommerce sales prediction

This project uses linear regression to explore which variables have the greatest impact on a company's sales. It is found that the length of membership and time in the mobile application influence the most. A predictive model is made with linear regression.

California coast housing prices prediction

This project uses a dataset of house prices based on various attributes such as location, number of rooms, and proximity to the beach in the state of California. Using machine learning techniques, we can build a predictive model for house prices based on their location.

Google Vision

In this code, we use Google's artificial intelligence to automate image labeling by identifying elements present in the image and generating a list of keywords for each one. This can be useful for improving image searchability, organizing large datasets, enhancing SEO for websites, and optimizing content for machine learning applications.

Distribution map of road incidents in CDMX

In this heat map, we visualize the distribution of traffic incidents in Mexico City over specific time intervals. It allows us to identify potentially dangerous areas that require attention and helps prevent future accidents.

Low-Resolution Image Scaler

This web application enhances the quality of low-resolution images using advanced artificial intelligence techniques. Through a simple interface, users can upload an image and obtain an upscaled version with improved detail and sharpness.

Technologies used:

• Flask: Framework for web application development.
• PyTorch: Library for deep learning and neural network processing.
• Real-ESRGAN: AI-based model for image super-resolution.
• PIL (Pillow): Image processing and manipulation in Python.
• NumPy: Data handling and matrix operations for optimized processing.

How it works:

1. The user uploads an image to the platform.
2. The image is processed with the Real-ESRGAN model, based on RRDBNet, to enhance resolution and details.
3. The enhanced image is generated and displayed for download.

This project is ideal for restoring old images, scaling graphics in multimedia applications, and improving image quality in low-resolution environments.

US 2020 elections map

In this interactive map, we visualize the results of the 2020 U.S. elections using a color-coded system, allowing us to check the percentage of votes for the Republican and Democratic parties.