sadman rahman

At Penta Global Ltd., I supported research and experimentation on multimodal audio and image models, along with technical documentation. I also contributed to voice privacy work exploring how to anonymize speaker identity while preserving utility.

My projects include an ongoing effort to extend the Rano speaker-anonymization paper to detect AI-generated speech while preserving speaker privacy. I’ve also developed a Bengali–English cross-lingual information retrieval system combining lexical and semantic retrieval with query translation and named entity recognition.

At Islamic University of Technology, I studied Computer Science and Engineering and worked on projects spanning AI applications, pose tracking, and games. I’m especially interested in research and development in machine learning and AI, particularly training AI models.