Radmantis

Continuous monitoring of individual fish

This is the monitoring half of the platform. Radmantis uses machine-learning inference to detect, identify, and characterize individual fish underwater — by species, size, sex, health, parasite load, and other biosecurity-relevant traits. In a U.S. Department of Energy–supported study, our vision platform reconstructed species composition with 99% accuracy and detected fish with 95.5% reliability, even in dense, fast-moving groups. Core methods are under review for international patent protection.

Closed-loop control, driven by on-device inference

Real-time fish handling requires a closed control loop: the system must sense, classify, and act on a fish within the same moment, with no room for a round trip to the cloud. That's why inference runs locally, on AMD Kria™ adaptive system-on-modules — the model executes in the water and drives the device's physical response directly. Cloud latency would break the loop; on-device edge computing is what makes autonomous, real-time action possible, even at remote sites with limited power or connectivity.

Built to safeguard fish and waterways

Biosecurity only works if the technology reaches the water. We build modular devices that deploy across aquaculture facilities, hydropower intakes, and open waterways with minimal customization — standing up to harsh conditions, unreliable power, and limited communications, often all at once. Pairing edge computing with semi-autonomous management, each unit protects its site independently and reliably.

From monitoring to safeguarding with autonomous action

This is where monitoring becomes safeguarding. Radmantis translates real-time classification into physical action — sorting, guiding, or selectively removing individual fish by chosen criteria — so threats are mitigated and healthy populations protected without manual handling. We prove and refine these capabilities at our own commercial Recirculating Aquaculture System (RAS) in Northwest Ohio, raising trout to market size and expanding the platform with each new robotic device.

Real-time monitoring and remote control for fish farms

Radmantis gives facility managers real-time visibility into their fish populations and the ability to manage them remotely. By replacing manual handling with automated, data-driven monitoring, our systems reduce stress and disease, improve fish welfare, and support the shift toward sustainable, land-based aquaculture.

Helping migratory fish move safely past dams

Radmantis develops AI-guided fish passage solutions that help migratory species navigate dams and hydroelectric infrastructure safely. By identifying and guiding fish in real time, our technology protects vulnerable populations and supports habitat connectivity while allowing energy operations to continue.

Field-deployable detection and selective removal

Radmantis builds field-deployable systems that combine advanced species classification with selective extraction of invasive species. By accurately distinguishing target invasives from native populations, our technology enables precise, low-impact removal that protects native ecosystems and biodiversity.