Geospatial AI for Indigenous data sovereignty.

Self-hosted software and federated machine learning that let each Nation monitor cumulative effects and turn satellite, sensor and field data into verifiable land and water intelligence, while its raw data stays under its control.

ArchitectureSelf-hosted, federated ML
Backed byMila, Quebec Artificial Intelligence Institute
CompanyIndigenous-led
Data governanceDesigned around OCAPĀ®

Approved one at a time. Felt all at once.

Canada assesses development project by project. The land carries every road, cutline and well pad together, for decades.

In Yahey v British Columbia (2021), the BC Supreme Court found that cumulative industrial disturbance breached Treaty 8. By 2018, 85 percent of Blueberry River First Nations' claim area lay within 250 metres of industrial disturbance. Cumulative effects are now a legal test, and Nations need evidence they own.

REIA Cumulative Effects
Satellite image of the Christina Lake area, Alberta, in 1985: forest crossed by a few roads The Christina Lake area, Alberta, in 1989 The Christina Lake area, Alberta, in 1994 The Christina Lake area, Alberta, in 1999 The Christina Lake area, Alberta, in 2005 The Christina Lake area, Alberta, in 2009 The Christina Lake area, Alberta, in 2017 The Christina Lake area, Alberta, in 2020 The Christina Lake area in 2025, crossed by roads and industrial clearings
Christina Lake, Alberta, 54 by 35 km, 1985 to 2025. Landsat 5, 7 and 8, courtesy of the U.S. Geological Survey.

From orbit, a forest. In the public record, a grid.

Drag across 16 by 10 km near Christina Lake. On the right, the satellite image. On the left, what public records add: seismic lines, pipelines, roads and wells, and the land within 250 m of them.

Sentinel-2 satellite image of a 16 by 10 km area near Christina Lake, Alberta, September 12, 2025: forest crossed by roads, well pads and industrial clearings.
The same area with public seismic lines, pipelines, roads and wells drawn over the image, and the land within 250 m of them shaded.
Public recordsSentinel-2, September 12, 2025
99%of the land is within 250 m of mapped disturbance
over 99%is within 500 m
under 1%chance two points are still in one unbroken patch
Real data. Christina Lake area, Alberta, 16 by 10 km. Copernicus Sentinel-2 imagery; seismic lines, roads, pipelines and wells from public Government of Alberta and AER layers; REIA engine at 10 m.
See REIA Cumulative Effects
Beta, December 2026

From boundary file to referral response.

  1. Install

    Install REIA on a laptop or server the Nation chooses. After setup it runs without an internet connection.

  2. Add your territory

    Load your boundary from a file you already have. It's stored only in your local data file.

  3. Run a check

    Get disturbance, connectivity and change results, with every source and setting recorded.

How the platform works

Your data stays under your control

You choose where it runs.

Your laptop, your server, fully offline if you want.

You own everything it produces.

Data, results and anything derived from your territory belong to the Nation.

Nothing is shared without consent.

Nothing from a Nation trains a shared model without written consent that can be withdrawn.

Leaving is built in.

Export everything, or delete it all with one command.

TerminalExample
$ reia nation delete example-nationType the Nation name again to confirm: example-nationRemoved nations/example-nation/reia.gpkgRemoved nations/example-nation/evidence/ (214 files)Removed nations/example-nation/cache/Done. No data for example-nation remains on this machine.
Beta, December 2026The delete command, with invented example data.

REIA is designed around the First Nations principles of OCAPĀ® (ownership, control, access and possession). These are commitments we can be held to, not a certification. Read our data sovereignty commitments.

Research, results expected spring 2027

Learning together, without pooling data

We're researching how several Nations can improve a shared Earth observation model while each Nation's raw imagery, labels and sensitive layers stay on its own hardware. Only small model updates, called adapters, move between machines. We're also testing how a Nation can withdraw its contribution, and measuring privacy with established attack tests. Results will be published, including what doesn't work.

This is active research in Mila's Founders' Residency. Privacy test results are statistical evidence, not proof.

Federation Illustration: three Nations keep their data on their own machines. Only small model updates travel to a shared model, and one Nation can withdraw. Example adapter, 4 MB adapter, 4 MB adapter, 4 MB Shared model Nation A raw datastays here Nation B raw datastays here Nation C raw datastays here

What REIA doesn't do

  • It doesn't decide anything for a Nation. Every output is evidence for a person to check, never a finding of wrongdoing.
  • It doesn't map sacred, burial or harvesting sites.
  • It doesn't send a Nation's data to a cloud AI service.
  • It doesn't use Traditional Knowledge as training data.
  • It doesn't name any Nation publicly without written Council approval.

Built by members of the communities it serves.

Ruth Anna Ritchotte

Co-Founder and CTO

Saulteaux, Muskowekwan First Nation. Leads engineering and machine learning, and built REIA's cumulative effects engine.

Mishra Cameron

Co-Founder and COO

Syilx, Okanagan Indian Band. Earth Sciences and Data Science at the University of Victoria. Leads operations, community partnerships and the earth science.

Mila

Winner of Mila's 2026 Indigenous Pathfinders in AI pitch competition, and part of Mila's Founders' Residency.

About REIA

Start with your territory.

Every enquiry goes directly to our founding team.

For Nations

Talk to us about your territory

A first pass on public data takes weeks, not years, and shows what a full build adds. The data is yours from day one, in writing, including the right to take it and leave.

For funders and partners

Fund geospatial AI for Indigenous data sovereignty

We share our technical overview, our roadmap and the research plan behind federated learning for Indigenous land data.

For researchers

Partner on the science

Remote sensing, privacy-preserving machine learning, hydrology and restoration ecology. Community ownership of the resulting data is a condition of every partnership.

Careers

Join the team

Remote sensing, data engineering, ecology and community-facing roles. Indigenous applicants are strongly encouraged, especially from northern and on-reserve communities.

Goes straight to us, nowhere else. We reply within two working days. You can also email info@reia.ca.