Radical Numerics has launched Omnii Defense, an AI-powered edge model designed to detect both natural and synthetic pathogens. The system operates offline, requiring only a laptop and a USB gene sequencer, which allows for deployment in remote areas without internet access.
The technology is part of the Omnii family, based on the Omnii-1 multimodal genome language model. Unlike traditional tools that compare samples against known libraries, Omnii-1 identifies genetic patterns to recognize pathogens, including those that are mutated or AI-generated.
In internal benchmarking against the bioinformatics tool DIAMOND, Omnii Defense achieved a pathogen recall rate of 95.8% at a 0.5 probability threshold, compared to 78.6% for the competing tool. The model is capable of identifying divergent sequences and novel viral strains.
The company, founded by Eric Nguyen and Michael Poli, offers the technology via an API for large-scale screening alongside the offline version. Radical Numerics is currently collaborating with national laboratories, including Lawrence Livermore, to advance its biodefense capabilities.
