Part of the Cybersecurity team working on artificial intelligent tasks, mainly using state of the art NLP and Anomaly Detection models. Additionally I have been working on the development of an scalable Machine Learning Platform (ML Ops) in AWS.
Teaching 'Programming I' in C for first-year computer science students & Professor of 'Fundamentals of Programming and Communications' in Python for fourth year students of Electric Engineering.
Here I worked for multiple projects from different industries. For exmaple I fulfilled all data science duties for a leading Japanese car manufacturer firm. Additionally, I deliver a highly accurate object detection system to validate the drugs distribution from one of the most important Spanish drug distribution companies.
Slashed the maintenance cost of Datawarehouse by creating an advanced ETL (based on Azure Integration services) pipeline to efficiently store and analyze the incoming heterogeneous data.
EfficientDet is a state-of-the-art object detection model that stands out for its efficiency and accuracy. It utilizes a scalable architecture and compound scaling method, which optimizes the balance between resolution, depth, and width of the network, leading to improved performance with lower computational cost.
Pointer Network (Ptr-NN) is a neural network architecture, based on sequence2sequence models, which is able to learn the conditional probability of an output sequence with elements that are discrete tokens corresponding to positions in an input sequence.
KitNET is an online, unsupervised, and efficient anomaly detector. A Kitsune, in Japanese folklore, is a mythical fox-like creature that has a number of tails, can mimic different forms, and whose strength increases with experience. Similarly, Kit-NET has an ensemble of small neural networks (autoencoders), which are trained to mimic (reconstruct) network traffic patterns, and whose performance incrementally improves overtime.
Utility for One Piece Treasure Cruise phone game. Export your One Piece characters from your box automatically from screenshots. The project uses object detection + contrastrive learning to match the game characters faces to the original anime characters.
GUI payload chainloader for Nintendo Swtich. The purpose of Argon NX is to stay immutable, so you can always inject it, without caring about other payloads getting updated (Always use ArgonNX for TegraSmash, TegraGUI, TrinkedM0...).
Contributed to fix low quality code and supporting Amazon Titan Express model as a ChatModel
I also contributed to many open source projects: For example torchvision, SageMaker SDK, transformers, JAX... See the list of all my PRs clicking the 'See more...' button.
I'm Guillem, a passionate Machine Learning Engineer. My expertise lies in Natural Language Processing, Computer Vision and Anomaly Detection, which I have successfully applied in various industries.
At Siemens, where I currently lead as a Machine Learning Enginer, I am involved in the Cybersecurity team, developing AI solutions. My work includes developing/training ML models and creatring an scalable MLOps platform in AWS. This role has enabled me to reduce cyber-attack risks through NLP and unsupervised anomaly detection techniques. Beyond Siemens, I have shared my knowledge and passion for technology as an Associate Professor at the University of Lleida.
My approach to ML and Data Science is not just about creating advanced algorithms, but also about ensuring that these solutions are practical, scalable, and adaptable to changing business needs. This focus is reflected in my commitment to MLOps and infrastructure development.
