Students from the University's Master's Degree in Big Data Science share their experience at the work End of Year Master's Degree
The final module of the Master's Degree University in Big Data Science focuses on developing a project that solves real problems presented by companies or institutions with which we have agreements partnership. This allows the student to gain significant internship experience in the field of data and big data science.
Álvaro Carracedo, Jesús López and Jefferson Osorio are part of a team of students who are developing their TFM with the manager Artificial Intelligence and data at Telefónica.
"The work de Fin de Master's Degree focuses on the assessment comparative of various generative Artificial Intelligence (GenAI) imaging technologies, more specifically diffusion models. This study seeks to identify and select models, then define and apply quantitative and qualitative metrics that evaluate the quality of the generated images."
Through the implementation of a testing environment and comparative analysis, we aim to identify the strengths and weaknesses of each model, culminating in conclusions and recommendations on its effectiveness and possible improvements.
Paula Sanjuan and María Simal are carrying out their TFM at signature international technology consultancy service NTT Data.
The clinical narrative defined as the free and detailed text that physicians write in patients' medical records, is an often underutilized source of information. These records can contain data critical for medical research and clinical decision making, from understanding rare or unknown diseases to early detection of syndromic outbreaks. However, the free format in which these texts are presented makes it difficult to take advantage of them, losing valuable information in the process.
"Our project focuses on assessing the validity and refining open source language models (LLMs), particularly Mistral, for the recognition of symptomatology in free text and the assignment of their respective codes according to ICD-10 standards. This task is crucial for converting clinical narrative into structured and useful data . To improve the performance of these Generative AI models, we are exploring several refinement strategies: Prompt Engineering, Retrieval-Augmented Generation (RAG) or Fine Tuning."
Paula and Maria seek to transform the way clinical narrative is used. "By structuring these data effectively, we facilitate their analysis and exploitation, contributing significantly to the medical research and the clinical internship ."
Santiago Rosell, Data Scientist at BBVA and Giulio Brevi, EU Central Operations Forecasting Manager at Amazon are developing a project under the supervision of BBVA.
The work uses back-testing techniques to evaluate the accuracy of the evaluation models of companies by comparing the intrinsic value of stocks with their historical market value.
"The goal of our End of Master's Degree work is to identify patterns and trends to improve future valuations and help investors make more informed decisions. This approach allows to adjust evaluation models and optimize investment portfolios, facilitating a more efficient and adaptive management to market conditions," says Giulio.
On the other hand, Pablo Legerén and Mario Lamas do their TFM by developing a project at partnership with EY WavespaceThe company, a Center of Excellence in innovation and transformation through the application of AI, generative AI and analytics from data, is part of one of the world's leading professional services companies. His work is focused on developing a recommender system for the online platform Airbnbintegrating techniques and knowledge from Master's Degree to solve complex Big Data problems.
"We started by generating a database using web scraping, a method we learned in the subject Collection Techniques from data. Then, we performed a thorough cleaning of the dataset, essential for the quality of the data, using data Preparation and Cleaning skills. Finally, we employed natural language processing (NLP) techniques to enrich the dataset with user feedback information, improving the accuracy of the recommender system," says Mario.
Finally, Paloma Duarte, an analyst at Mutua Madrileña, based her work on the prediction of the rate of leave customers through Machine Learning techniques.
The project has as goal to identify patterns and key variables that influence customers' decision to cancel their policies. This early prediction allows to anticipate and implement personalized and efficient retention strategies, thus improving customer loyalty and business sustainability.
"The procedures learned in the subject Cleaning and Preparation of data, such as the detection and elimination of outliers, imputation of missing values and transformation of variables, have been crucial to ensure the quality and integrity of the data, which in turn improves the accuracy and reliability of the predictive model ," Paloma comments.
