Egoitz Gonzalez
About me
Hi there, I'm Egoitz!
I am currently finishing my Master's degree in Artificial Intelligence at the University of Amsterdam. Prior to this, I obtained a double Bachelor's degree in Physics and Electronic Engineering, which also equipped me with a solid foundation in math and programming.
My primary interests are Machine Learning, Deep Learning, and Computer Vision and other applications such as VR/XR. I am eager to apply my expertise in designing, developing, and implementing AI-based solutions that create meaningful impact. I am also enthusiastic about engaging in new projects and continuously learning new tools and methodologies.
Education
-
MSc AI @ UvA, Amsterdam, NL 2022-2024
A research Master's degree with strong mathematical and applied technical skills that includes many different projects that encourage teamwork. I am more interested in the fields of Deep Learning and Computer Vision.
-
BSc Physics + Electronic Engineering @ EHU, Bilbao, ES 2016-2021
Experience
-
Teaching Assistant @ UvA, Amsterdam, NL 2023
Teaching assistant in Computer Vision course. Assignment preparation, assistance and grading.
-
Teaching Assistant @ UvA, Amsterdam, NL 2023
Teaching assistant in Computer Vision course. Assignment preparation, assistance and grading.
-
Python Developer @ sherpa.ai, Bilbao, ES 2021-2022
I worked on a Federated Learning Platform, including the development of the application backend and the communication between distributed agents. Developed in python and using multiple AWS tools.
-
Student Intern @ Cambrian Intelligence, ES 2021
Analysis of medical data and development of predictive models for respiratory syndrome disease emergencies.
Projects
-
Uncracking the Bagel repo report poster
Anomaly detection on 3D point cloud data using point-voxel diffusion. We use denoising diffusion models to learn the distribution of healthy data, i.e. learning to generate samples without anomalies. Then, given an anomalous sample, we apply some steps of noise to the point cloud data and reconstruct its healthy version using the diffusion model. Taking the difference between the original anomalous sample and the reconstructed healthy sample we can detect the anomalies. This project is part of my master's degree in AI.
-
VAE-VDM repo blogpost supplement poster
We explore representation learning using probabilistic diffusion models. We use an architecture for variational autoencoders that uses a variational diffusion model (VDM) as the decoder and a custom CNN as the encoder. We train the VDM conditioned on a latent representation learned by the encoder and analyze to what extent the encoder is able to learn meaningful representations in the latent space. This project is part of my master's degree in AI.
-
[Re] ReLIC: Bias in Captioning Models repo OpenReview
Reproducibility study on quantification of bias amplification by image captioning models. We reproduce the results of this work using their LIC metric and extend their work to see whether it may be applicable for other attributes. This project is part of my master's degree in AI.
-
DeeLeMMa repo
Simple toolkit to build modular deep learning models. It was built from scratch based on pytorch or keras interfaces. Through this project, I was able to gain practical knowledge and experience in implementing the fundamentals of DL models, as well as resolving unexpected issues that can arise during the coding and model training. It was developed as part of my final bachelor's degree project.