Carlos' Portfolio
AI & Data Science Projects
Project Portfolio
A collection of academic and personal projects. Take a look at my recent work!
You are what you do, not what you say you'll do - Carl Jung
Parallel Stencil Computation
High-performance, hybrid MPI+OpenMP implementation of a 2D stencil heat diffusion solver, explicitly optimized for the Leonardo DCGP cluster (Intel Sapphire Rapids).
Biases in T2I Generation: An Analysis of Qwen-Image
Investigation into the inherent biases present in modern Text-to-Image (T2I) generation models. Utilizing a local installation of the Qwen-Image model via ComfyUI, we generated batch outputs across various prompts to identify representational harms.
Nexus.CS
Interactive educational platform that uses a knowledge graph to structure learning paths for Computer Science and Artificial Intelligence. It maps broad domains into a visual hierarchy of interconnected nodes, linking high-level concepts down to specific, detailed study modules.
Analysis of RL Algorithms for a Simulated Hill Climb Racing Agent
A comparative framework for analyzing reinforcement learning algorithms in a custom Hill Climb Racing environment. It implements PPO, DQN, and Expected SARSA, supporting neural network, linear and polynomial function approximators.
Underwater Fish Detection, Tracking, and Classification
A robust computer vision pipeline for underwater fish analysis. It uses a YOLOv8 model for stable, real-time tracking and classification. Features multiple processing modes, including buffered real-time analysis and high-accuracy offline filtering, making it a flexible tool for marine biologists and researchers.
Probabilistic Deep Forest
Novel implementation of the "Probabilistic Deep Forest (PDF)", a model that enhances Deep Forest to handle noisy, real-world data. It solves the critical issue of uncertainty loss during layer-by-layer prediction by using Probabilistic Random Forests as its core estimators.
From Context-Free to Natural Language Grammars: A Genetic Approach
This project explores the use of Genetic Programming to evolve a population of context-free grammars, to approximate the structure of real grammars (such as English or Spanish).
LyrIAcs: AI-Powered Lyric Generation
Web-based songwriting assistant that leverages ML and LLMs to help users compose song lyrics.
Cluster Creation and Testing with VirtualBox VMs and Docker Containers
Creation and performance comparison of virtualized versus containerized clusters: 1. Cluster of Virtual Machines (VMs) with Oracle VirtualBox. 2. Containerized cluster with Docker containers.
Cardiovascular Disease Data Analysis and Prediction
Analysis centered around the prediction of presence of cardiovascular diseases. Data is preprocessed and various analysis are performed over it. Multiple models are applied: K-Means Clustering, Random Forest, Logistic Regression and Natural Splines. Finally, a result overview and comparison is provided.