Mohammad Omar Khursheed

I am an Applied Scientist at Amazon, where I work on developing efficient machine learning systems for speech recognition. My research focuses on creating robust and fair deep learning architectures for resource-constrained environments, with a particular interest in the intersection of model efficiency and speech representations.

My work explores how to optimally balance information preservation with constraints of memory and compute, developing solutions for real-world deployment challenges. Recent projects include designing dynamically slimmable transformers for keyword spotting and investigating fairness considerations in speech processing systems.

Research Interests

• Efficient Deep Learning Architectures
• Speech Recognition & Keyword Spotting
• Machine Learning Fairness
• Self-supervised Speech Representation Learning