Radmantis

Radmantis develops uncrewed systems that use robust machine learning inference to categorize and sort individual fish entering the device.

Key technologies for performing this task are currently under review for international patent protection in markets where aquaculture production has a significant presence. Fast on-device inference of a machine learning model is executed locally to classify features of particular interest. In aquaculture settings these might include individuals of a particular sex, nutritional status, size, reproductive status, disease incidence, parasitic load, or the presence of other acute stressors. Field-deployed devices might focus on population monitoring with species identifications, demographics, or predictors of fecundity.

We develop modular devices that can be implemented in a wide range of sites with minimal customization.

Common challenges include deployment under harsh environmental conditions, with limited access to reliable power, or insufficient communication infrastructure -- and often all three simultaneously. We balance these competing needs by coupling an edge computing paradigm with semi-autonomous management systems. The latter are able to perform complex machine learning inference in situ, and then based on classification with respect to the chosen criteria (e.g. species, size, sex, health), guide routines designed to optimize outcomes.

Radmantis operates a small-scale, commercial, indoor Recirculating Aquaculture System (RAS).

At the Agricultural Incubator Foundation in Northwest Ohio we raise trout to market size in a system that follows standard aqauculture practices. Serving as a practical proving ground for our fish management technologies, the system will expand to a multi-unit production system with the continued addition of robotic devices for automated task management.

Radmantis develops real-time solutions that allow facilities managers to monitor and remotely manage a facility's fish population

The system captures longitudinal changes in a range of morphological and behavioral phenotypes. Using a web-enabled dashboard, managers can monitor progress in development, track population health and welfare, and perform a range of functions such as selective harvesting, recognizing the emergence of threats to biosecurity, and extracting problem individuals in a timely fashion. We develop control functions for Recirculation Aquaculture Systems (RAS) with industry partners, and we currently seeking collabroations on expanding our technology into coastal marine netpen operations.

Our company develops field-deployable systems that combine advanced species classification with selective extraction of invasives.

Implemented at natural choke points, our AI-based devices are designed to recognize harmful invasives and to divert them into secure holding. Our automated designs advance a technological ‘immune system’ for habitats under attack, one that can detect the presence of nuissance individuals and perform a species-selective capture.

Our solutions are designed to mitigate ecological damages to fish populations at water intakes and diversion structures.

Radmantis uses its patented, uncrewed technologies to ward off deadly entrainment of fish at water management structures, such as utility intakes. Automated and adaptable technological solutions recognize a danger to target fish and aims to deter them from sources of entrainment. By adding a layer of smart technologies. AI-based solutions enhance, rather than aim to replace existing structures.