Muhammad Usama

I am the AI Initiative Team Lead at Braindeck Inc. in Seoul, where I lead work on speech and voice AI for Korean-language users with dysarthria. This includes a custom automatic speech recognition model built for dysarthric speech, integrated into a real-time avatar communication platform, and an end-to-end pipeline for diagnosis, rehabilitation, and communication assistance. My work at Braindeck also covers multimodal emotion recognition from speech, facial expressions, and biosignals.

I received my Ph.D. in Electrical Engineering from the Korea Advanced Institute of Science and Technology (KAIST) in 2026, advised by Prof. Dong Eui Chang in the Control Lab. I also hold an M.S. in Electrical Engineering from KAIST and a B.E. in Electrical Engineering from the National University of Sciences and Technology (NUST), Pakistan.

My research interests span large language models and multi-agent systems, speech and voice AI, reinforcement learning, signal integrity for high-speed memory systems, and machine learning applications in medicine. My doctoral thesis focused on latent representation learning for anomaly detection and reinforcement-learning-based equalizer optimization in high-speed DRAM systems, carried out in collaboration with Samsung Electronics Device Solutions (DS).

My experience, education, and publications and patents each have their own page, and a PDF version of my CV is also available. You can also find my work on Google Scholar and GitHub.