Biography

João is a Research Scientist at Autodesk, with interests in autoregressive generative modeling and information retrieval, and a focus in multimodal settings. In particular, João has been focusing on devising approaches to render inference of language models more efficient, especially in settings where models should be given very large input prompts. Moreover, João has focused on robustness aspects of various model classes across various settings such as testing for the ability of language models to find knowledge in user-provided content, or on how to train encoders able to operate in multiple data modalities and across varying languages. João obtained a PhD from the Institut National de la Recherche Scientifique – University of Quebec in Montreal, where he focused on applications of generative modeling and metric learning to voice biometrics settings. His goal was to improve the state-of-the-art in tasks such as speaker verification and spoken language identification, while also improving the robustness of such approaches against spoofing attackers that relied on synthetic data.