Free space optical communications has a number of advantages for the transmission of high bandwidth data. However, for many applications, its use is limited by the precise alignment tolerances required to maintain a reliable link. In this dissertation, the design and construction of a free space adaptive optical interconnect demonstrator system is reported, in which the transmitted beam was actively steered to compensate for misalignment. The target application considered was a board to board interconnect for use within computer systems. Beam steering was performed using binary phase gratings displayed on a Spatial Light Modulator (SLM), for which an ‘off the shelf’ ferroelectric Liquid Crystal on Silicon (LCOS) micro display was used here. The gratings were generated in hardware, as an integrated part of the SLM driver, using a novel and compact implementation for which the details are described. This was capable of generating gratings at high frame rates, and applied a scrolled addressing scheme to ensure DC balance of the pixels whilst maintaining an uninterrupted optical path. Data transmission through a bulk optical relay containing this SLM was successfully demonstrated at 2.5Gbps. The transmitter and receiver modules were custom built for these experiments using an 850nm multi mode VCSEL and PIN photodiode, driven by standard telecommunications components. Losses due to the optical components, SLM, grating diffraction efficiency and scrolled addressing scheme totalled between 15.1 and 17dB. These corresponded well with the values estimated for the components, and with further device optimisation it was expected that they could be substantially reduced. Beam steering at the receiver plane was achieved over a 6.4x6.4mm range with a resolution of 25μm. This was sufficient to track a detector with a coupling loss of no more than 0.05dB. The feasibility of adaptive alignment correction in a free space optical interconnect, using a liquid crystal SLM for beam steering, was demonstrated through a series of experiments. The considerations relevant to extending this system for multiple parallel channels are discussed.