Not too long ago the strongest 10um silicate emitting AGN currently known was serendipitously discovered behind the Magellanic Clouds (MCs). SAGE1C J053634.78−722658.5 (SAGE0536AGN) lies at z=0.14 and has a smaller than expected galaxy mass, an unusually intense silicate emission line, as well as a negligible amount of far-IR (infra-red) emission and therefore a lack of emission from star-formation. Unusual/Extreme sources such as this provide us with the unique potential to see a window into galaxy evolution not often observed, and in this case a chance to study the AGN environment without confusion from star-formation regions. Last year we discovered SAGE0534AGN behind the MCs at z=1, which also had silicate emission, though not as extreme, and a lack of emission from star-formation, making this a potential link to SAGE0536AGN’s extreme state. Another similarity between these sources was that they were both originally believed to be evolved stars. The MCs are an excellent place for multiwavelength studies due to the plethora of observations from X-ray to radio that have been taken of them. Through a machine learning technique called t-distributed stochastic neighbourhood embedding we have already discovered several candidates for higher-z SAGE0536AGN look-alikes. The large survey speed of the Nancy Grace Roman Space Telescope has great potential for uncovering similarly rare but key sources that will help us understand both the universal and individual evolutionary paths.