The exponential rise of generative AI technology has created unprecedented demand for GPUs, which current supply is struggling to keep up with. To address this AI developers need to use their GPU capacity more efficiently. Traditional sequential programming does not fully utilize the computational power available from modern multi-core GPUs. This results in underperformance, where the potential of the hardware is not fully exploited, leading to slower processing times and increased operational costs.