Cai Selvas Sala

About Me

I'm an Artificial Intelligence student at the Universitat Politècnica de Catalunya (UPC), passionate about leveraging Machine Learning, Deep Learning, Data Science and Data Analysis to solve complex problems. With a solid foundation in programming languages and AI, I strive to create impactful solutions in the field of Artificial Intelligence.

Current Studies

• Bachelor's degree in Artificial Intelligence (2022 - Present): Universitat Politècnica de Catalunya (UPC), Barcelona, Spain.
  • Highest honors in Reinforcement and Unsupervised Learning, Programming and Algorithms I.
  • Mentor of newcomers.
  • Current GPA: 8.73/10.

Experience

• Research Intern at Computer Vision Center (2025 - Present): Researching Machine Unlearning techniques for Vision-Language Models (e.g., CLIP, SigLIP, LLaVA), focusing on ensuring data privacy while minimizing the impact on overall model performance.
  • Rosa Sensat fellowship awardee (2025).

Skills

• Programming Languages: Python, R, SQL, C/C++, MATLAB.
• Main Skills: Machine Learning, Deep Learning, Computer Vision, Reinforcement Learning, Unsupervised Learning, Natural Language Processing (NLP), AI Algorithms, Data Science, Data Analysis, Mathematics, and Statistics

Hackathons and Events

• [Aina Hack 2024] Winner project (LaIA): LaIA is an AI assistant for Catalan administration that simplifies access by accepting files and images as inputs, supporting internet searches, and generating informative text and video outputs to guide users through administrative processes.
• [HackUPC 2024] InterSystems challenge: Creation of a program that suggests travel destinations and points of interest by leveraging user prompts, image recognition, and Retrieval-Augmented Generation (RAG) using embeddings on the InterSystems IRIS Data Platform.
• [Challenge IA i Auditoria 2024] Winner project: A project to predict the inventory impairment using advanced AI models such as autoencoders (NN), advanced time series algorithms, EBM for explaining the results, word embeddings, RAG to interpret the results and more. We emerged to create a model user friendly in order to help auditors with their work.
• [Datathon FME 2023] NTT Data challenge: A project that forecasted the next year’s required purchases for a set of hospitals using time series analysis.
• [HackUPC 2023] Winner of Bunge’s challenge: Developed a program that processes requests in natural language to extract information from large datasets.

Languages

• English (C2 Level - Cambridge C1 Advanced Score: 202).
• Spanish (Native).
• Catalan (Native).

Past Studies

• Technological Baccalaureate (2020 - 2022): Institut Escola Industrial, Sabadell, Spain.
  • Grade: 9.1/10 (with honors), 12.44/14 (Selectivitat/PAU).