AI geoscience for critical mineral discovery.
DepoDart is a next-generation Mineral Discovery Intelligence platform that combines Mineral Prospectivity Mapping, AI, and integrated geoscience data to significantly improve the probability of discovering economic mineral deposits.
Three ways to work with DepoDart.
Exploration Projects
These are short, medium, or long-term exploration projects in which mineralization targets are identified through the integration of artificial intelligence and geoscience. Each project is carefully tailored to the geological context of the area of interest, enabling the accurate delineation and ranking of prospective mineralized zones.
The primary deliverables are high-resolution 2D and 3D geochemical prospectivity maps. In addition, DepoDart generates a range of complementary products, including lithological maps, structural geology interpretations, and geophysical variable models.
A key differentiator of our approach is explainability. Every prediction is accompanied by a narrative interpretation that describes the geological, geochemical, and geophysical factors contributing to the result, providing transparency and confidence in the targeting process.
Results can be delivered in a variety of formats; however, the most powerful option is an interactive software platform that enables users to visualize, explore, and analyze all associated data and interpretations in a single environment.
One of the platform's most innovative features is its adaptability. Users can customize and extend its functionality through natural-language prompts, allowing the system to generate new visualizations and analytical tools on demand based on user-defined requirements.
Interactive Data Preprocessing Tools
This software platform is designed to visualize, transform, and reformat geological datasets. It provides a comprehensive suite of statistical data-quality and preprocessing tools that can be applied either manually or through automated workflows.
For large datasets, users can configure and execute automated processing pipelines by selecting from a range of predefined actions. In addition, the platform enables users to create custom preprocessing operations through natural-language prompts.
These prompts are automatically translated into executable code, generating new tools tailored to the user's specific requirements. This capability allows geoscientists and data professionals to rapidly extend the platform's functionality without traditional software development, creating customized workflows that adapt to the unique characteristics of each project.
Free Online Regional Target Identification
This AI-powered mineral exploration platform leverages publicly available geological, geochemical, geophysical, and remote sensing data to identify prospective mineralization targets at a regional scale.
The application assists exploration teams during the early stages of target selection by generating two-dimensional prospectivity maps that delineate areas with the highest probability of hosting elevated concentrations of one or more commodities. In addition to target maps, the platform provides supporting geological insights and contextual information to facilitate exploration decision-making.
The system is continuously updated as new public datasets become available, ensuring that prospectivity models incorporate the latest information from government surveys, geological agencies, and other public data providers.
Users can also enhance the platform's predictive capabilities by uploading their own proprietary datasets through the web interface. These custom data sources can be integrated with public information, resulting in more refined and project-specific targeting outcomes.
Four deliverables. Every pilot.
Prospectivity surface
Color-coded raster (GeoTIFF) showing ore-forming probability per pixel across the area of interest, exportable at any resolution.
Ranked drill targets
Prioritised CSV and PDF report of top drill locations with probability scores, mineral type, estimated depth, and confidence tier.
Uncertainty map
Companion raster quantifying model uncertainty per pixel — highlights where confidence is high vs. where more data would help most.
Data provenance report
Full audit trail showing which input layers drove each recommendation — critical for defensible exploration campaign planning.
One model. Every element.
The platform is commodity-agnostic. A single model run can score multiple commodities simultaneously without retraining. Current active targets include:
- Copper
- Gold
- Lithium
- Cobalt
- Nickel
- Silver
- Uranium
- Iron
- Manganese
- Zinc
- Rare Earth Elements
- Platinum Group
From data handover to drill targets in days.
A DepoDart pilot is a focused, low-commitment engagement. You share a dataset covering your area of interest and we return a prospectivity surface with ranked drill targets, uncertainty bounds, and full data provenance. Here is exactly what happens at each stage.
Data Sharing
Day 1You share a geoscientific dataset covering your area of interest — any format, any vintage, any coordinate system. Common inputs include airborne geophysics, soil geochemistry, drill logs, geological maps, and satellite imagery. There is no minimum dataset size and no required preprocessing. An NDA can be executed before any data transfer. DepoDart uses your data solely to generate the deliverables for your pilot and does not retain or share it.
- Any geoscientific data format
- Area of interest boundary
- Target commodity or commodities
- Optional: known deposit locations for model calibration
Ingestion & Normalisation
Days 1–2DepoDart ingests your dataset and resolves all format conflicts, coordinate reprojections, schema mismatches, and coverage gaps automatically. Every input layer is quality-checked and registered to a common spatial reference. You receive a data receipt confirming what was ingested, what was excluded, and why — so there are no surprises in the output.
AI Model Run
Days 2–5The platform trains machine learning models on the geological signatures of known mineral occurrences within your dataset, then applies those signatures across the full area of interest. If no known deposits are present, the model uses regional analogue data. The model cross-references all input layers simultaneously, surfacing non-obvious spatial correlations that are impractical to detect with manual methods.
QC & Geoscientist Review
Days 4–7Model outputs are reviewed by DepoDart's geoscience team before delivery. We check that the prospectivity surface is geologically coherent, that target locations correspond to defensible feature combinations, and that uncertainty estimates are calibrated. We flag any areas where additional data would meaningfully improve model confidence.
Delivery
Days 5–14 (depending on dataset volume)You receive four deliverables: a prospectivity surface (GeoTIFF), a ranked drill target report (PDF + CSV), an uncertainty map (GeoTIFF), and a full data provenance report. All files are delivered via secure transfer. We walk through the outputs together and answer questions about the methodology, individual targets, and next steps.
- Prospectivity surface (GeoTIFF)
- Ranked drill targets (PDF + CSV)
- Uncertainty map (GeoTIFF)
- Data provenance report
Three-stage platform pipeline.
Data Fusion
DepoDart ingests and normalises multi-source geoscientific data regardless of format, coordinate system, or age. Drill logs, geophysics, geochemistry, satellite imagery, and structural maps are resolved into a unified spatial model automatically — no manual preprocessing required.
- Any coordinate system
- Legacy format support
- Conflict resolution
- Automated QC
AI Prospectivity Mapping
Our models learn the geological signature of known mineral deposits within your dataset, then apply that learned pattern across the full area of interest. Every point on the map receives a ranked mineralisation probability score, producing a continuous prospectivity surface at district scale and below.
- Supervised ML
- Multi-commodity
- District to deposit scale
- Probability surface output
Ranked Drill Targets
Every output includes a prioritised list of drill targets with confidence tiers, estimated depths, mineral type classifications, and uncertainty bounds. Full data provenance is included so your geologists can audit and defend every recommendation — no black-box outputs.
- Ranked priority list
- Uncertainty quantification
- Estimated depth
- Full provenance
Any format. Any vintage. Any coordinate system.
Clients deliver what they have. DepoDart resolves format conflicts, reprojections, and schema mismatches internally. There is no minimum dataset size or required preprocessing.
Geophysics
- Airborne magnetics (XYZ, GDB)
- Gravity surveys
- EM (ground/airborne)
- IP/resistivity
- Radiometric
Geochemistry
- Rock samples (CSV, XLS)
- Soil / sediment samples
- Assay tables (any schema)
- Lithogeochemistry
- Stream sediment
Geological
- Drill logs (LAS, CSV, DLOG)
- Geological maps (SHP, GDB)
- Lithology logs
- Structural interpretations
- Core imagery
Remote Sensing
- Multispectral (GeoTIFF)
- Hyperspectral
- SRTM/DEM elevation
- Landsat / Sentinel
- SAR derivatives
Common questions.
What if I have very little historical data?
Sparse data is expected. The model adapts to dataset density — smaller datasets produce prospectivity surfaces with wider uncertainty bounds, which is an honest reflection of what the data supports. We include the uncertainty map in every delivery so your geologists know where confidence is high versus where more data collection would help most.
Do you work with data from multiple projects or just one area?
A focused pilot covers a single defined area of interest. For clients with multiple project areas, we run separate pilots per area or discuss a broader engagement. There is no limit on area size — from a single tenement to a regional district.
What happens to our data after the pilot?
Your data is used solely to generate the deliverables for your pilot. DepoDart does not retain, share, or use client data for training models on other clients' projects. Data handling terms are covered in the pilot agreement before any transfer takes place.
Can we run additional commodity targets on the same dataset?
Yes. A single processed dataset can be scored for multiple commodities simultaneously or sequentially. If you want to add a commodity target after the initial pilot, we can re-run the model at a reduced cost since the data is already ingested.
Why exploration teams choose DepoDart.
Not a black box
DepoDart is transparent and explainable. It automatically surfaces the connections between each prediction and the inputs that drove it, so the result is easy to justify. Opacity is one of the central problems mining has with AI, and it is the part we deliberately set out to solve.
A close working relationship with the client
We work alongside you throughout the project so the output meets your requirements. The model is shaped around your data, your geology, and the questions you actually need answered.
A reliable model, unlike anything else on the market
Whatever your experience has been with other AI approaches, this one is different. The architecture is intrinsically distinct from anything else available and incorporates every piece of information at once, finding interactive patterns at a higher level.
Multivariable prediction
DepoDart does more than produce a prospectivity map for a single target. It returns a multi-level map covering the full set of variables across your land.
One pilot. Real data. A map you keep.
Share a dataset, and we return a prospectivity surface over your area of interest within days — with full data provenance included. No long-term commitment required.