SO-101 Policy Training
Built at the EuroTech x HKTE Hackathon: trained ACT and SmolVLA imitation-learning policies on a SO-101 robotic arm.
Built at the EuroTech x HKTE Hackathon: trained ACT and SmolVLA imitation-learning policies on a SO-101 robotic arm.
Research on novel diffusion-based architectures for large language models. Exploring more efficient text generation through length prediction mechanisms.
Graded as the best exam project for the Bocconi graduate course 20879 Language Technology.
Benchmarking explanation methods for CNNs in medical vision. This project evaluates different saliency methods for understanding how neural networks make decisions in medical image analysis.
Graded as the best (tied) exam project for the Bocconi graduate course 20878 Computer Vision & Image Processing. Published by Bocconi Students for Machine Learning.
Exploring semantic similarity retrieval methods maximizing BLEU score. Advanced NLP techniques for measuring and improving semantic understanding in text.
Super efficient algorithm for solving generalized Wordle puzzles. Implements optimal guessing strategies using information theory and probability.
This website! A minimal, clean design built from scratch with pure HTML, CSS, and JavaScript. Features dark/light mode and responsive design.