Master's Thesis:
Building Better Barriers: Evaluating Threats and Crafting Guidelines to Secure AI Code Generators
This thesis delves into the emerging security risks posed by data poisoning attacks on large language models (LLMs). Through rigorous experiments with state-of-the-art models like Phi3, CodeT5, and CodeGen, it uncovers critical vulnerabilities, including cross-industry effects and the "bleeding" phenomenon where poisoned outputs transcend domains. The findings emphasize the urgent need for robust, resource-efficient defenses, such as prompt-based systems, to safeguard the integrity of LLMs. This work contributes to advancing AI security research, addressing real-world challenges in industries relying on LLMs.

Bachelor's Thesis (only available in Spanish):
Desarrollo de Wemon: Una red social innovadora construida con Flutter y APIs REST en un equipo multidisciplinar.
This thesis presents a comprehensive exploration of software engineering methodologies applied to the front-end development of Wemon, a social network tailored to the Information Society. It highlights the structured approach to converting conceptual designs into user-focused app screens, leveraging REST APIs for state management. Developed collaboratively in a team of 6-8 members, this project showcases the importance of teamwork, user story refinement, and iterative development in building scalable, modern applications.