Efficient signal detectors are rather important yet challenging to achieve satisfactory performance for large-scale communication systems. This paper considers a non-orthogonal sparse code multiple access (SCMA) configuration for multiple-input multiple-output (MIMO) systems with recently proposed orthogonal time frequency space (OTFS) modulation. We develop a novel low-complexity yet effective customized Memory approximate message passing (AMP) algorithm for channel equalization and multi-user detection. Specifically, the proposed Memory AMP detector enjoys the sparsity of the channel matrix and only applies matrix-vector multiplications in each iteration for low-complexity. To alleviate the performance degradation caused by positive reinforcement problem in the iterative process, all the preceding messages are utilized to guarantee the orthogonality principle in Memory AMP detector. Simulation results are finally provided to illustrate the superiority of our Memory AMP detector over the existing solutions.