Maritime Autonomous Neural
Triage and Inference System

MANTIS is a ruggedized, rack-mounted compute appliance designed to be installed aboard maritime surface combatants and expeditionary platforms: presenting no fixed target and no single point of failure. It ingests standard ISR data formats from any unmanned platform the vessel receives, performs AI inference autonomously on full-motion video, synthetic aperture radar imagery, and signals data, and delivers curated intelligence products to watchstanders and combat systems in near-real time. Raw sensor data is processed where it is collected. Only the compressed intelligence product traverses the satellite link.

Time to actionable intelligence compresses from minutes to seconds. The ship retains full exploitation capability in EMCON and denied-communications conditions. Satellite bandwidth previously consumed by raw ISR feeds is freed for voice, command-and-control, and other warfighting functions. A single operator can oversee feeds that would currently require several dedicated watchstanders.

The processing core is built around commercial AI accelerator hardware: GPU-class inference processors offering high throughput-per-watt for sustained inference workloads. The NVIDIA Jetson AGX Industrial family provides the target performance envelope. These processors support mixed precision inference, enabling efficient execution of the heterogeneous AI workloads MANTIS must handle, from convolutional neural networks for video analysis to transformer-based classifiers for radar imagery.

For thermal management, MANTIS uses a closed-loop liquid cooling system designed to interface with the ship's existing chilled-water infrastructure, providing deterministic thermal performance independent of ambient compartment temperature and allowing the processing hardware to operate at sustained maximum throughput without thermal throttling.