Aims. We aim to locate previously unknown stellar clusters using the VISTA variables in the V a Lactea Survey (VVV) catalogue data. Methods. The method fits a mixture model of Gaussian densities and background noise and uses the expectation maximization algorithm to pre-filtered near-infrared survey stellar catalogue data; it was developed by the authors for the UKIDSS Galactic Plane Survey (GPS). Results. The search located 88 previously unknown candidates, most of which are embedded stellar cluster candidates, and 39 previously unknown sites of star formation in the 562 deg(2) covered by VVV in the Galactic bulge and the southern disk.