In massive machine-type communication (mMTC) systems, a large number of machine-type devices sporadically transmit small packets with low rates. By exploiting the sporadic activity of machine-type devices, we can cast the device activity detection problem as a compressed sensing based multiuser detection (CS-MUD) problem.  In this talk, we propose a novel algorithm for joint channel estimation and device activity detection using recent advances in the approximate belief propagation (BP) based Bayesian inference paradigm. The algorithm leverages the double sparsity resulting from the sporadic user activity and beam-domain representation of sparse multipath massive MIMO channels. To render the proposed algorithm fully automated, we also devise an expectation-maximization (EM) procedure that learns all the required hyperparameters online.