Uganda is home to over 52 local langauges. However, many of these lack adequate speech datasets. This project aimed to address this gap by collecting crowdsourced speech and text data for diverse African languages. Our goal was to leverage this data to research innovative architectures and deep learning algorithms for multilingual NLP systems (including Speech, NMT, Q&A, LM). These systems would be robust to variations in accents. Ultimately, Waxal strives to make NLP systems more inclusive with the development and inclusive of richer image prompt and transcribed datasets for Ugandan languages.