A considerable amount of what we know about galaxy composition and evolution comes from modelling spectral energy distributions (SEDs), and by establishing scaling relations derived from the computed properties. Since the time of the first successful deep surveys with HST, the size of and information gathered by galaxy surveys has grown exponentially, owing to the advent of new facilities, particularly in the IR-sub-mm regime. This, in return, has created a new challenge, namely a necessity for quick and reliable routines to process these large amounts of data. This motivated us to develop Stardust, a novel, fast and versatile, panchromatic SED fitting algorithm that quickly and reliably models the emission from stars, AGN, and infrared dust emission. We apply our code to a sample of ~ 5000 star-forming galaxies at z < 5, drawn from the deepest multi-wavelength cosmological fields, and produce one of the most statistically significant and homogeneous samples of infrared parameters to date. I will report the emerging trends that capture the evolution of the far-IR physical properties of star forming galaxies over the last 12 billion years, including the evolution of the dust to stellar mass ratio, of the dust mass function, and of the gas fraction. I will also discuss the discovery and the nature of the 'gas-giants', a puzzling population of galaxies that stand out of the established scaling relations, displaying remarkably elevated gas fractions and low star formation efficiencies. Finally, I will describe the unique capabilities of Stardust for future applications, and highlight our ongoing panchromatic study of high-z, faint sub-mm galaxies detected by ALCS, a large imaging ALMA program of the most well-known lensed cluster fields.