Gravitational lensing is a rare phenomenon which requires the exact alignment between a distant source and a massive foreground galaxy. The light from the source is bent around the galaxy forming spectacular rings, arcs and multiple images which have captured the public imagination since their discovery in 1979. This thesis focuses on searches for lensed systems where the background source is an active galactic nucleus (AGN) - a galaxy powered by their central supermassive black hole (SMBH). Gravitational lensing of the most distant and luminous subclass of AGN, known as quasars, provide a unique way to study the universe. They have been used to constrain the Hubble constant (H0), probe the dark matter content of galaxies and uncover the proposed co-evolution between SMBHs and their hosts. Until recently, these studies have been hindered by low sample sizes. However, increased research into automated approaches such as machine learning, combined with a wealth of data from new wide-field surveys, will enable us to dramatically increase the lensed quasar population. In this thesis we develop a number of searches leveraging classic colour selection as well as supervised machine learning techniques. We target quadruply imaged and high-redshift lensed quasars using photometric data from DES, WISE and Pan-STARRS, together with Gaia astrometry. Candidates were confirmed with follow-up spectroscopy, resulting in the discovery of several new systems. A search for radio-loud AGN was conducted using newly released VLASS data and candidates were prepared for high spatial resolution radio confirmation. We discuss the serendipitous discovery of an exciting lensed radio-loud galaxy with extreme high ionisation UV line emission and damped Lyα absorption. We finally outline a deep learning approach to automate the visual inspection stage of our lens searches in preparation for the release of new surveys such as LSST and Euclid.