Quasars are highly biased tracers of the large-scale structure and therefore powerful probes of the initial conditions and the evolution of the universe. However, current spectroscopic catalogues are relatively small for studying the clustering of quasars on large-scales and over extended redshift ranges. Hence one must resort to photometric catalogues, which include large numbers of quasars identified using imaging data but suffer from significant stellar contamination and systematic uncertainties. I will present a detailed analysis of the photometric quasars from the Sloan Digital Sky Survey, and the resulting constraints on the quasar bias and primordial non-Gaussianity. The constraints on $f_{\rm NL}$, its spectral index, and $g_{\rm NL}$, are the tightest ever obtained from a single population of quasars or galaxies, and are competitive with the results obtained with WMAP, demonstrating the potential of quasars to complement CMB experiments. These results take advantage of a novel technique, 'extended mode projection', to mitigate the complex spatially-varying systematics present in the survey in a blind and robust fashion. This work is a new step towards the exploitation of data from the Dark Energy Survey, Euclid and LSST, which will require a careful mitigation of systematics in order to robustly constrain new physics.
