Common questions about AI mineral exploration.
Everything you need to know about AI mineral prospectivity mapping, what data you need, and what to expect from a DepoDart pilot.
What is DepoDart?
DepoDart is an AI platform for mineral exploration. It ingests geological, geophysical, and geochemical data, learns the signatures of known mineral deposits, and produces ranked drill targets and prospectivity maps for an area of interest. Our team brings together geologists, data scientists, and software engineers, and we work directly with exploration companies to turn their data into actionable guidance for discovery and mine development.
What is AI mineral prospectivity mapping?
AI mineral prospectivity mapping uses machine learning models to score every grid point within an area of interest based on its probability of hosting a mineral deposit. Traditional prospectivity mapping requires a geologist to manually define a deposit model, specifying which geological, geophysical, and geochemical criteria indicate mineralisation. AI prospectivity mapping inverts this: the model learns the deposit signature from the data itself, by studying the feature combinations near known mineral occurrences, then applies that learned pattern across unexplored terrain. The output is a raster highlighting the areas most likely to host undiscovered deposits. DepoDart automates the full pipeline, from raw data ingestion to the delivery of a ranked prospectivity surface.
How does DepoDart find drill targets?
DepoDart trains supervised machine learning models on large datasets. The models learn correlation patterns among geophysical, litho-geochemical, and structural features. It then uses this knowledge to estimate the chemical concentration of several minerals at every point of a geological grid. The output is a multi-raster in which each pixel contains the estimated concentration of several chemical elements, each with a mineral-system classification. Full data provenance is included so geologists can audit which input layers drive each recommendation and defend the target selection.
What data does DepoDart need to start a pilot?
Any combination of geoscientific data covering your area of interest works. Common inputs include airborne or ground geophysics (magnetics, gravity, EM, IP, radiometrics), rock, sediment, and soil geochemistry, and geological structures. There is no minimum dataset size. DepoDart automatically normalises mixed formats, coordinate systems, or mixed resolutions, so you can deliver what you have. If you know deposit locations or mineralised intersections within the area, including them significantly improves model calibration. If no known occurrences are present, the model uses regional analogue data from publicly available geological surveys.
Can I export analysis results from DepoDart?
Yes. DepoDart delivers results along with an interactive application for visualising, reshaping, and exporting them to GeoTIFF, GDB, XYZ, CSV, and any other industry standard. Ranked drill targets are provided as PDF reports and CSV files that include coordinates, probability scores, mineral-type classifications, and uncertainty bounds. Source data provenance is included so your team can audit, reuse, and rerun the inputs in future modelling. You retain full ownership of every output we deliver.
Which critical minerals can DepoDart target?
DepoDart is commodity-agnostic. The model learns deposit signatures from data, so it adapts to any mineral system. Current active commodities include copper, gold, lithium, cobalt, nickel, silver, uranium, iron, manganese, zinc, platinum group elements, and rare earth elements. A single model run can score multiple commodities simultaneously without retraining, and the platform cross-references all input layers to identify overlapping anomaly signatures indicative of multi-commodity mineralisation. For clients exploring battery-critical minerals such as lithium, cobalt, and nickel, the platform models the full suite of battery metals.
How long does a DepoDart pilot take?
A focused pilot, from data handover to full deliverable delivery, typically takes five to fourteen days, depending on dataset volume and format. Ingestion and normalisation take one to two days. The deployment of the model takes two to five days, including tuning. Geoscientist QC review takes one to two days. The complete output includes a prospectivity surface (GeoTIFF), a ranked drill-target report (PDF and CSV), an uncertainty map, and a full data-provenance report.
How frequently is the data updated?
Our underlying geoscientific datasets are refreshed continuously as new public surveys are released and as we expand into new regions across Canada, Australia, the United States, and Europe. For active client engagement, we rerun the model whenever you provide new inputs, such as recent drill logs, new geophysical surveys, or updated geochemistry, so your prospectivity surface stays current with the latest field information. There is no fixed update interval; the cadence is driven by what your project needs.
How accurate are AI-generated prospectivity maps?
Accuracy depends on data density, data quality, and the number of known occurrences available for model calibration. In well-sampled areas with multiple known deposits, DepoDart consistently ranks historically productive zones in the top quartile of the prospectivity surface. In data-sparse areas, accuracy is lower, and this is transparently communicated via an uncertainty map. Each pixel carries both a probability score and a confidence interval.
How accurate is DepoDart?
We measure accuracy, project by project, using cross-validation against known mineral occurrences within your area of interest. The model is trained on a subset of known deposits and scored on the held-out remainder; the predicted prospectivity is then compared against actual occurrence locations. In well-sampled regions, our models consistently rank productive zones in the top quartile of the prospectivity surface. Every output ships with an uncertainty layer so geologists can see where the model is confident and where data is sparse, and you can independently validate the result against your own legacy drilling before committing to a new program.
How does AI exploration compare to traditional geological targeting?
Traditional geological targeting relies on a geologist defining a deposit model and manually cross-referencing datasets to find where criteria converge. This is expert-dependent, time-intensive, and limited by how many layers a human can hold in memory simultaneously. AI prospectivity mapping learns the deposit model from the data, cross-references all input layers in full, and scores every point on the map in a fraction of the time. It surfaces non-obvious spatial correlations and subtle combinations of anomalies across multiple datasets that are impractical to detect manually. The result extends the geologist's intuition rather than replacing it: the geologist interprets and validates the AI output using their domain knowledge.
Is my data secure with DepoDart?
Your data is treated with strict confidentiality. We sign a mutual non-disclosure agreement before any data exchange, store all client data in encrypted, access-controlled environments, and never share or aggregate proprietary client data into our public models. Each project is processed in isolation, and you retain full ownership of both the input data and the resulting deliverables. On request we can also process your data inside your own infrastructure for engagements with specific compliance requirements.
What does a DepoDart pilot cost?
Pilot pricing depends on the size of the area of interest, data volume, and the number of commodity targets. Contact us at [email protected], and we will follow up with a scoped proposal within one business day. A pilot is a low-commitment, high-signal engagement: the deliverables are yours regardless of whether you proceed to a longer engagement. We are focused on exploration in Canada and the United States and are onboarding pilot clients in Q2 2026.
