The Earth’s properties, composition and structure ahead of a tunnel boring machine (TBM) should be mapped for hazard assessment during excavation. We study the use of seismic-exploration techniques for this purpose. We focus on a seismic system for soft soils, where shear waves are better and easier to interpret than compressional waves, as has been shown over the last decade. The system is intended to be deployed on the machine’s cutter head, with a few seismic sources and sufficiently many seismic sensors to tackle spatial variability and noise characteristics. An important property of the newly developed system is its ability to process data with very little human interaction. Images need to be available in near real time, without human interactions slowing down the imaging process. This can be achieved by employing Full Waveform Inversion, which minimizes the difference between modeled and observed data. Because this method may suffer from local minima in the cost function if the data lack low-frequency information, we employ a dedicated seismic source that can generate sufficiently low frequencies for the relevant length scales. With data acquired in a number of field settings that mimic realistic TBM configurations, we show that the designed seismic system can successfully look ahead of the TBM and offers a valuable capability to support decision-making during tunnel excavation.