Projects
Interesting projects worth mentioning.
A multimodal learning framework that learns a joint representation of language and music, based on the CLIP model. Trained on a large-scale dataset of music and tags using ViT for mel-spectrograms and BERT for language. Won "most marketable solution" at the 1st Sound of AI Hackathon.
Daily arXiv paper digest on Telegram, covering NLP (cs.CL), Computer Vision (cs.CV), Audio (cs.SD), and Multimedia (cs.MM). Automatic ranking and TL;DR summarization of the latest deep learning publications.
Multi-year project developing NLP tools for detecting and correcting stereotypes, gender bias, and ableism across multiple languages (Italian, French, Spanish, and more).