AI for
Earth's living systems.

Land trust managers, watershed councils, and tribal natural-resource programs are collecting eDNA at scale. They send a sample. They get back a list of species. They can't tell whether the result is healthy, recovering, or declining — because every lab runs its own assay, primer, taxonomy, and schema, and there's no standardized reference to compare against.

For a research lead, an agency program manager, or a natural-resource department, the practical effect is the same — every study is an island. Findings don't compose. Comparison across regions or years is manual, slow, and contested.

PHYLO sits downstream of any assay or sensor — hardware-agnostic by design. It pulls together the relevant public reads, harmonizes preprocessing and taxonomy across institutions and schemas, and produces a standardized corpus that's both directly queryable and the prerequisite training data for genomic foundation models.

The defensibility is in the corpus and the multi-source pipeline, not in any single model. The output is a queryable view of life — what is there, in what abundance, in what relationships, and how it is changing across studies, regions, and time.

The end-to-end pipeline is in production on real-world sample data, with results that hold up under expert review. Standardization scoring is on the near-term roadmap.

PHYLO runs on Google Cloud, accelerated through NVIDIA Inception MEMBER, with infrastructure credits via AWS Activate. The work has been validated by domain experts at the eDNA Collaborative, Conservation X Labs, and senior operators in commercial eDNA. We work with the research community, not against it: open data, proprietary platform.