A nonprofit building the missing index of science.

Who we are

COMPASS ID, Inc. is a nonprofit research organization headquartered in La Jolla, California. We were founded on a simple observation: the world produces millions of scientific papers, but there is no map of what all that research actually covers. Knowledge concentrates on a handful of well-studied species, regions, and questions, while vast areas have no indexed research, and no one can see the pattern.

We build the tools to see it. The literature itself is just an enormous pile of unstructured text — searchable for words, but not something you can ask real questions of. Our method is to have AI read every paper and attach the same accurate, structured metadata to each one: what was studied, and where. Once every paper is structured the same way, the pile becomes a database — queryable in plain language, and because location is one of the fields, mappable, showing the empty spaces as clearly as the crowded ones.

Why we exist

In conservation, what is not studied is easy to overlook. A species no one has studied is one no one is protecting, and a funder cannot close a gap they cannot see. We began with biodiversity because the stakes are immediate — but the same method applies to disease, agriculture, earth science, and beyond. Our goal is a public good: a shared, trustworthy picture of what humanity knows, freely available to the people who can act on it.

The work exists because of a gap in the scholarly record itself. A paper’s metadata says who wrote it, where they are employed, and which journal published it — but nothing says where the research happened. That is true of geography, and equally true of where in a body, a cell, or a genome a study sits. Every one of those locations has to be read out of the paper. That is what we build: the layer the record never had. In practice each paper gets a study location read from its own text and resolved to coordinates, which we then set against the countries where its species are known to occur. The coordinates are machine-read and checked on a sample, and the range comparison is at country level, so both limits are stated wherever the figures appear. We call this layer, and the reasoning built on it, spatial intelligence: a machine working out where things are and how places relate, applied to the literature itself. Read more →

How we work

Each tag COMPASS ID makes uses entities from external references (IUCN, GBIF, MeSH), cites a quote from the paper, and is reviewed by a separate model. We publish our methods and measured accuracy openly; the tagging pipeline itself has not yet been peer reviewed.

COMPASS ID’s literature classification runs on Anthropic’s Claude models, supported by an award of API credits through Anthropic’s AI for Science program.

The platform is hosted on Microsoft Azure with credits from Microsoft for Nonprofits, and Google for Nonprofits provides our Google Workspace and Ad Grants.

Leadership

COMPASS ID was founded by Dr. Chris McCarthy, who leads the organization’s research and platform development. For partnership, collaboration, or governance inquiries, please contact us.

Selected publications