I'm Aitor Egurtzegi, a researcher in computer vision at TOELT-LLC working on data science, data analysis, AI, and Machine Learning with a background in Computational Linguistics and NLP.
I work on deep learning for computer vision. With my colleagues, I develop metrics for image comparison and fine-tune super resolution models for thermal images. Additionally, I also work on data compliance, data risk assessment and data protection.
During my PhD I worked with language models and language corpora, I ran EEG, eye tracking, and online experiments, and I used machine learning methods and Bayesian statistical modeling to understand how language works.
Recent research
Recently I have been working on computer vision problems coupled with data protection and compliance. More specifically, I've been looking at the use of different metrics for image comparison and how their results vary depending on the type of image one works with. While co-supervising a master's thesis, we developed a new metric that brings better results when comparing images of galaxies, given the stark contrasts such images have: the changes in such images only happen in a small part of it that is of relevance, and our algorithm can disregard the void background while focusing on the planets and stars.
Moreover, I've also been working on super-resolution methods for thermal images in order to better identify features in human faces: given a low-quality image, these deep learning models can convert it into a high-quality image with the goal of improving landmark detection and facial recognition in thermal images.
Furthermore, I have also been involved in data protection and compliance, drafting data risk assessments, data sharing agreements, joint controllership agreements, and ethics protocols following Swiss and GDPR regulations.
PhD research
During my years at University of Zurich, I worked on various projects at different departments. In one such project I spent my time doing data analysis with a focus on Machine Learning and Bayesian hierarchical regression. Particularly, I tackled the status of word categories in a two-fold manner: on the one hand by preprocessing, analyzing, and visualizing MEG data looking for spatial underpinnings of said categories and on the other hand by developing NLP pipelines working with state-of-the-art language models to analyze said word categories using clustering methods, running cloud computing jobs using a Unix-based system.
Moreover, in another project I also did experimental work to collect eye tracking and EEG data and analyze it in a Bayesian framework for low-resourced languages that are typologically different from the most common spoken languages.
The key point I've found out is: the language you speak affects how you perceive the world around you.
For instance, speakers of ergative languages such as Basque, Georgian, or Hindi spend less time looking at agents and more time looking at the overall picture to faster grasp the action, since their grammar requires them to know this faster than speakers of nominative languages such as English, German, or Spanish.
Additionally, these language-based differences in perception are also appreciated in the brain activity, with the latter languages needing a stronger activation of the neuronal population because they entertain more, competing structures.
Eye tracking
Alpha activity
For more information on my work, see some of my PhD publications in Brain and Language and PLOS Biology.
About me
I hold a PhD in comparative psycholinguistics from University of Zurich. I'm interested in how language is produced and understood, equally from a computational and a neurological level.
Previously, I studied a double master's degree in Computational Linguistics and NLP from The University of Malta and Saarland University, and I graduated in Linguistics from The University of Barcelona.
I'm curious by nature. I love traveling, hiking, discovering new countries, learning random geography facts, playing chess, basketball, working out and watching various sports, from college football to basketball, Australian rules, soccer, or cricket, among others.
Contact
If you wish to reach out, feel free to use the buttons on the left side of the page to email me, connect with me on LinkedIn, or DM me on LinkedIn or Twitter (I do check my email and LinkedIn more frequently than Twitter).