David Pérez Carrasco

Member of the AI & ML Research Group at Universitat Pompeu Fabra. Exchange student at University of British Columbia

I am a passionate and driven ML/AI researcher with a year of experience in the field of reinforcement learning. As an intern at the AI & ML research group at Universitat Pompeu Fabra, I developed and evaluated scalable versions of Q-learning and Z-learning algorithms within MDPs and LMDPs, focusing on efficiency and robustness in Grid World environments, achieving up to a 4000-fold speedup over default implementations.

Recently, I completed my Bachelor’s Thesis, “Efficient Algorithms for Linearly Solvable Markov Decision Processes”, under the guidance of Anders Jonsson, head of the research group. My thesis introduces novel embedding techniques for mapping MDPs into LMDPs, enhancing approximation precision by 99.23% over existing baselines. Additionally, I evaluated traditional MDP models against LMDP models, demonstrating superior performance, while developing efficient methods to enhance RL agents’ decision-making within the LMDP framework. This research marked the completion of my degree in Mathematical Engineering in Data Science from UPF, where I graduated second in my class.

My interest in ML/AI was sparked during my exchange at the University of British Columbia, where I took courses on Machine Learning, Artificial Intelligence, Data Science, and Statistics. Fascinated by the potential and challenges of this field, I decided to delve deeper into it, leading me to gain hands-on experience with various deep learning (DL) and natural language processing (NLP) projects. My dedication to research is demonstrated by my accelerated academic progress, completing all coursework six months ahead of schedule. This allowed me to fully dedicate myself to research and remain readily available for exciting new opportunities in the ever-evolving realm of AI and ML.