New Zealand is a geodynamically complex region subject to diverse natural hazards. This archipelago has two subduction zones dipping into the mantle in opposite directions, with the capital city of Wellington situated above a potentially hazardous subduction-zone fault whose stress regime remains poorly understood, placing the city at risk of earthquakes and tsunamis. To investigate this hazard, the Earthquakes and Locking inVEstigation of Subduction (ELVES) project deployed 20 Ocean Bottom Seismometers between November 2023 and January 2025 offshore Wellington, in the locked zone of the Hikurangi subduction zone. In this work, we perform a preliminary analysis of the OBS data by detecting and classifying seismic waveforms using a deep-learning autoencoder implemented in the RUMBLE (Rapid Underwater Mass movement and Benthic Listening Exploration) software platform to identify and extract natural (e.g., earthquakes, submarine mass wasting events), anthropogenic (e.g., ships) and cetacean signals. Those signals will then be analyzed in isolation to focus on events of geological or biological origin. In particular, this study will focus on mass-wasting events, which are difficult to distinguish in seismically active areas and whose occurrence patterns remain unknown, yet they pose a significant hazard. In addition, cetacean detections and tracking may inform studies of their movement patterns off the coast of New Zealand.