How IRIS works

Most models pick a signal. IRIS reads the system.

A deposit doesn't live in one dataset. It expresses itself across geology, geophysics and geochemistry, and only surfaces when the data is read as a single geological system, not layer by layer. That is what IRIS does. And it is setting a new standard for how explorers find what others miss.

How a mandate works

You bring the data. If public geoscience datasets are missing from your picture, we fill the gaps. We deploy a fresh IRIS instance and train it on your data. It learns your geology, your known deposits, the specifics of your ground. You then receive favourability maps you can act on immediately, followed by a detailed report from our geologists that explains the results and brings in their exploration expertise. Every deployment is exclusive — nothing shared, no foundational model in the background. Full data sovereignty, by design. Once your instance is calibrated, new survey data integrates rapidly and your targets update to reflect it. Your model keeps pace with your exploration.

The IRIS approach

01 — A geology-aware deep learning model

No limits on data type or format

Maps, samples, rasters, vectors

IRIS encodes all exploration data into a single model. Geological, geochemical, geophysical, and more. Whatever your project holds, it goes in. The model learns the optimal combination of signals and the relationships between them.

02 — No data density bias

Finds targets where others stop

Including under-explored ground

Targeting tools cluster where data concentrates. Drill holes call for more drill holes. IRIS surfaces meaningful targets in data-sparse areas too. The blank spots, the under-explored ground. That is often where the real misses are.

03 — Built for your geology

Fresh instance, every project

No foundational model. No shared data.

A Cu porphyry in Gaspésie and one in the Cordillera share a deposit type. Not a local signature. IRIS trains a fresh instance for each project and each geological context, learning the specifics of your ground. Nothing shared. Full data sovereignty, by design.

04 — Geologically coherent results

Results you can defend

Structures, contacts, mineralization patterns

IRIS targets align with real geological structures, lithological contacts, and known mineralization patterns. Results a geologist can interpret, present to their team, and drill with confidence.

The proof is in the rock

IRIS is validated across lithium, gold, titanium and copper projects — greenfield and brownfield, multiple jurisdictions.

70–95 % Reduction in
target footprint
~35 % Precision uplift
vs baseline ML targeting
66.9 % of known Au deposits in IRIS top 3%
no grade data used (Abitibi backtest)

Multi-scale lithium search

IRIS independently prioritised the same zones as the client's exploration plan — with greater spatial precision. The drilling results confirmed the prediction. On the property, drill holes inside IRIS targets intersected mineralization. Those outside did not.

Adaptive re-targeting & discovery

When an initial drill hole could not be completed as planned, IRIS's pre-defined targets allowed the drill hole to be immediately repositioned within the same project area. Semi-massive mineralization was encountered directly on the target.

Where IRIS earns its place

  • Targeting a district-scale land position before committing ground.
  • First campaign on a grassroots property with no prior drill history.
  • Finding what’s been missed on an advanced-stage property.
  • Pre-campaign target ranking before a drill season starts.
  • 3D drill targeting* for resource expansion or new-zone discovery.
  • Mineral-potential inventories at jurisdiction scale (government agencies).

* 3D drill targeting is in active development. We work with selected client-partners on favourable terms. Get in touch to find out where we stand.

Talk to our team