The Arcminute Microkelvin Imager (AMI) is a dual-array radio interferometer sited at Lord's Bridge, near Cambridge. Although it was designed specifically for studying galaxy clusters via observations of the Sunyaev-Zel'dovich (SZ) effect, it is also an ideal instrument for Galactic science. This thesis describes science programmes investigating both Galactic objects and galaxy clusters that I have carried out with AMI.

A new data analysis pipeline is described which has been developed to allow the automated processing of data taken by AMI in drift-scan mode, pointing the telescope at a fixed azimuth and elevation and observing the sky that drifts past. This is a very efficient mode for large-scale surveys, but the different character of the data has required innovative algorithms for effective processing.

The AMI Galactic Plane Survey uses drift-scanning to cover the northern Galactic plane between |b| < approximately 5 degrees. It is the first Galactic plane survey at cm-wave frequencies to achieve crucial mJy-sensitivity levels at arcminute-scale resolution over a wide area, and as such provides a unique opportunity to investigate hitherto unusual objects such as ultra- and hyper-compact HII regions. I describe my work on the survey strategy and its implementation and on some of the science I have extracted so far including the follow-up of candidate hyper-compact HII regions.

The recently-released Planck satellite results include the largest catalogue of SZ-selected clusters of galaxies to date. I describe the AMI follow-up programme to observe the clusters within the AMI observation limits, and present the first results from the programme including an interesting discrepancy between the cluster parameters according to AMI and Planck. Since the two instruments are observing the same physical process, this indicates a fundamental problem with the 'universal' pressure profile currently favoured for modelling clusters.

In an attempt to address the discrepancy, I use simulations to investigate the effect of allowing the shape of the pressure profile to vary. The derived parameter constraints are found to vary when clusters are not simulated and recovered with the same model; the effects are dependent on angular size, worsening for larger clusters. I also assess the potential for using AMI data to constrain the cluster shape parameters, and conclude that weak constraints on the shape parameters are possible with a careful choice of prior.