Buscador de Tesis y publicaciones académicas

Automating Alan's internal processes

Recalde Tarron, Martí

Deploying LLM-based AI agents in real operational environments is harder than just building them. This thesis documents the design, implementation and deployment to production of an agentic system at Alan, the leading health-tech scale-up in the French ecosystem. We built it to solve the problem of automating one of th...

Fairness impact of machine unlearning in skin lesion classification

Blanco López, Elena

This work investigates the interplay between machine unlearning and fairness in medical image classification, a largely underexplored area despite the growing importance of both topics. Using skin lesion datasets (MILK10k and Fitzpatrick17k-C) as case studies, we analyze how the removal of specific subsets of training...

Implementation of genomics' visas with Verifiable Credentials' protocols and information models

Mayoral Soler, Elena

El estudio del genoma humano es esencial para comprender el comportamiento humano, identificar la raíz genética de las enfermedades, desarrollar tratamientos médicos y explorar los orígenes y ancestros humanos. Para llevar a cabo este trabajo, los investigadores requieren acceso a conjuntos de datos (datasets) grandes...

Deep reinforcement learning-based joint handover and trajectory optimization for 5G-connected UAV

Bermúdez Granados, Marina

In recent years, there has been a notable rise in interest around autonomous aerial vehicles. In spite of their broad applications and their affordable deployments, these drones still require a reliable connection in order to successfully accomplish their assignments. Since private networks are expensive and unscalable...

Design and implementation of a fault-tolerant orchestration system for distributed execution of LLM models on a heterogeneous machine network

González Vidal, Marc

Si bien la ejecución distribuida de Large Language Models (LLMs) ofrece una vía para utilizar hardware con recursos limitados, la gestión de estos entornos dinámicos plantea desafíos significativos. Los frameworks actuales suelen carecer de la coordinación robusta necesaria para manejar redes heterogéneas, el node chur...

Transformer-based models for predicting queueing delays in network traffic

Havuc, Eren

Network operators rely on simulation tools for capacity planning and topology design, but the high computational cost of traditional discrete-event simulators makes them practically unscalable for large networks. Recent work has tried to replace the computationally expensive queuing component with neural networks (NNs)...

Diffusion-based head-and-neck CT synthesis under limited data

Davtyan, Tatevik

Computed Tomography (CT) is central to the diagnosis and treatment planning of many oncological diseases. Head and neck cancers, which involve anatomically complex and functionally critical structures affecting breathing, speech, and swallowing, require particularly precise imaging for accurate clinical management. How...

Study of the integration of a passive reactivity control system in a sodium- cooled fast reactor

Dumont, Alex

A stigmergy-driven multi-agent framework for intelligent task orchestration

Veli, Eva

Contemporary LLM-based orchestration platforms such as LangGraph and CrewAI enable rapid composition of agent pipelines, but share three architectural weaknesses. They lack a persistent record of peer performance across runs, expose no formal model of agent capabilities and task constraints, and treat compliance and et...

Suvremena laboratorijska dijagnostika alergija

Džajić, Anica

The aim of this review thesis is to present contemporary diagnostic methods in allergology, with an emphasis on the development and application of in vitro laboratory analyses, molecular diagnostics, and functional tests. This thesis aims to consolidate current knowledge on their role in the assessment of sensitization...