Monitoring earthquakes in the ocean is difficult due to limited access to seafloor sensors, noisy environments, and the high cost of long-term deployments. Distributed Acoustic Sensing (DAS) offers a promising solution. However, many DAS systems require “dark fiber” (unused cables), which are not always available offshore. In this study, we tested a method called multiplexing that allows DAS to run on active telecommunication cables without interfering with their normal use. In May 2024, we deployed this system on a cabled network offshore Oregon and successfully recorded 31 regional earthquakes, demonstrating high-quality data comparable to conventional DAS.To improve earthquake detection in noisy ocean environments, we also developed a machine learning workflow that cleans DAS signals, identifies seismic wave arrivals, and combines DAS data with broadband seismometers. Using this approach in Alaska’s Cook Inlet, we significantly improved signal quality and detected about twice as many seismic phase arrivals as before. Over six months, we created a detailed earthquake profile, revealing deep structures in the subduction zone. These results show that combining machine learning, DAS, and multiplexing can greatly improve offshore earthquake monitoring.