Environmental DNA metabarcoding is widely used to survey marine biodiversity, but in practice, most workflows still require samples to be frozen, shipped to a laboratory, sequenced on high-throughput platforms, and processed through bioinformatic pipelines that demand specialist expertise. Weeks to months typically separate sample collection from species identification, and field teams rarely see their own results before leaving a site. We built and field-tested a fully portable genomics platform that compresses this entire pipeline, from water sample to species list, into a matter of hours, in a system that fits in a single backpack and can be operated by any field biologist. Streamlined wet-lab protocols eliminate the need for a cold chain, enabling field extraction and amplification at ambient temperature. The system pairs nanopore sequencing with GPU-accelerated basecalling on a low-cost edge computer and a real-time taxonomic display that returns species identifications directly to field teams as reads are generated, without requiring bioinformatic expertise. A custom demultiplexing algorithm recovers multiplexed samples that would otherwise be lost to barcode misassignment under field conditions. We deployed this platform at two extremes of the infrastructure spectrum. Aboard a research vessel crossing the western Pacific during a deep-sea expedition, we ran real-time multi-marker eDNA metabarcoding at sea, detecting deep-sea vertebrate and invertebrate eDNA including taxa associated with hydrothermal vent systems as the ship transited between stations. We then deployed the identical system on a remote island with no power grid, no laboratory, and no connectivity, processing intertidal and air eDNA entirely off-grid using solar charging and battery power alone. We report what each deployment detected, the practical limits we encountered, and lessons learned, from thermal management and power budgets to what matters most when infrastructure is not guaranteed. We are also developing machine-learning approaches to classify reads that no reference database can identify, a common problem in undersampled environments like the deep sea. The result is a reproducible, infrastructure-independent blueprint for real-time marine biodiversity monitoring in settings that have largely inaccessible to molecular observation, deployable anywhere on earth, by anyone.

暂无