Advancing Low-Latency Gravitational-Wave Detection through Machine Learning

With the growing volume and complexity of gravitational-wave (GW) data, machine learning and artificial intelligence are shaping the analysis efforts in GW astronomy. The advances in modern hardware have enabled faster and more cost-effective computation, allowing machine learning frameworks to operate efficiently at low latency. Leveraging these developments, we have developed end-to-end ML pipelines, such as Aframe and AMPLFI, that are currently running live in the LIGO-Virgo detectors for compact binary coalescence detection and rapid parameter estimation. These pipelines are built within the broader ML4GW framework, providing a set of tools for gravitational-wave analysis and inference. Achieving low-latency performance is especially critical for multi-messenger astronomy, where rapid sky localization enables timely electromagnetic (EM) follow-up of GW events. Building on these developments, we are extending these pipelines toward efficient, low-latency searches for binary neutron star mergers, a key source for joint GW–EM observations. I will highlight progress in building such low-latency ML frameworks, discuss results from public searches on LIGO data, and outline future goals to strengthen existing pipelines by further advancing real-time detection, improving early-warning capabilities, and enhancing the prospects for multi-messenger astronomy.