Dr. Chowdhury specializes in designing Conversational AI models, primarily addressing complex challenges such as multispeaker interactions, nuanced multilingual and dialect variations, and code-switching, among various other intricate conversational dynamics. She is currently the leading the speech technology in Fanar - Arabic AI Large Language Model and also LPI on the QVoice project, which empowers speakers—both native and non-native of all ages alike—to learn spoken Arabic. Dr. Chowdhury has received numerous awards and grants, including the NVIDIA Academic Hardware Grant for her research in simulating human language learning capabilities using DNN-based language models, a study that was also conducted as a part of the TRAILs project, funded by PRIN MIUR. As a key contributor to the EU-funded projects SENSEI and PortDial, Dr. Chowdhury developed conversational models adept at understanding human conversation, facilitating automatic summarization and mental health screening. She authored over 60 peer-reviewed publications in top-tier conferences and journals and played an active role in the research community by organizing shared tasks, challenges, and workshops, as well as serving on the committees of top-tier conferences and special interest groups. She co-founded the Bangla Language Processing Community and MyVoice, a crowdsourced platform, designed to bridge the gaps between standard and dialectal Arabic resources.

• Speech Processing: Representation Learning, Self-Supervised Models, Atypical Human Interaction, Speech Discourse & Turn-taking, and Spoken Language Understanding.

• Natural Language Processing: (Large) Language Models and their task understanding capabilities, Benchmarking.

• Explainable and Inclusive Speech Technology:
  Dialectal and Accented Speech Recognition, Pronunciation Assessment, Children Speech Recognition, Multilingual Models.