Barq Minerals jointly inverts every geophysical signal — gravity, magnetics, EM, resistivity, IP — into one precise subsurface model. The platform every geologist can use. For lithium, copper, cobalt, nickel — the minerals the energy transition depends on.
Critical mineral deposits don't sit on the surface anymore. Our AI surfaces high-probability targets buried under cover, where legacy methods go blind.
Every prediction comes with calibrated uncertainty. Know not just where to drill — but how confident to be, and what each new data point will reveal.
Discovery used to take a decade. We compress it into months — and every drill makes the next smarter, compounding precision across the campaign.
Different surveys see different versions of the same Earth. Gravity sees density. Magnetics sees magnetization. Electromagnetics sees conductivity. Each one alone is ambiguous — infinitely many subsurface models can produce the same data. Joint inversion is the discipline that fuses them into a single, sharper picture. Below: every major method ever developed, and the synthesis we believe wins for critical minerals.
Force the gradients of two model fields to be parallel. Structural features align across surveys — faults, ore-body boundaries, lithology contacts — without assuming the physical properties themselves are correlated. The first joint-inversion idea that actually worked.
If you know the rock-physics law connecting two properties — Gardner's velocity-density, Archie's resistivity-porosity — bake it directly into the inversion. When valid, the constraint is brutally tight. The catch: you have to know the law. Mining geology rarely gives you that luxury.
One mathematical object. Three couplings. The Gramian stabilizer enforces correlation (linear), structural similarity (gradient), or petrophysical relations (transform) — depending on what you put inside it. The most powerful general framework ever devised for multi-physics inversion. Yellowstone magmatic system, sub-basalt imaging, Australian IOCG deposits.
Don't pick one best model. Sample all models consistent with the data. Transdimensional MCMC even lets the data decide model complexity. The output isn't a single answer — it's a posterior distribution over earth models, each with a probability. This is the only honest way to tell a driller "drill here, with 87% confidence."
Physics-informed neural networks encode Maxwell's equations, Newtonian gravity, and elastic wave propagation as soft constraints in the loss function. Multimodal fusion architectures (MP3D-NET, 2025) treat each survey as a different modality and learn the cross-coupling end-to-end. Solves PDEs in milliseconds, not minutes. The bottleneck that legacy software cannot break.
The mathematical machinery is universal. PET-MRI joint reconstruction in medical imaging uses structural similarity priors (Bregman distances, total variation) — identical mathematics to cross-gradient. 4D-Var data assimilation in atmospheric science fuses satellite, radar, and surface sensors via the same Bayesian framework. The deepest learning here: this is a solved problem in physics, an unsolved problem in product.
The five paradigms above are not competitors. They are complements. The first company to fuse all of them into a single, browser-native, AI-accelerated inversion engine wins the critical-minerals software category. That's the hypothesis Barq is built on.
After each drill hole, the posterior conditions on the new evidence. RL picks the next best location — sequentially, optimally, always.
Lightweight alternative to MCMC. Maintains a swarm of plausible earth models. Every drill recommendation comes with a confidence interval.
Magmatic Ni-Cu-Co sulfide, porphyry Cu, sediment-hosted Li brine, IOCG, VMS, lateritic Ni. Baked-in priors that legacy generic software lacks.
Each modality — geophysical raster, drill log, hyperspectral, assay text — encoded into a shared latent space. The Gramian operates here, not on raw voxels.
Maxwell's, Newton's, elastic-wave PDEs solved by neural networks in milliseconds. The bottleneck that makes legacy software take weeks.
The mathematical core. One algebraic object that unifies structural, correlation, and petrophysical coupling. Any survey, any pair, any time.
No legacy product combines all six. SimPEG has the math, no DL. KoBold has ML, no inversion. Leapfrog is visualization. Mineflow is geometry. Barq is the AI-native joint-inversion layer underneath all of them.
The five paradigms above, fused into a single browser-native engine. Toggle data sources to watch the model sharpen — every signal you add tightens the posterior. Drilling data triggers POMDP + RL constraints to lock in the next drill target.
Every project ships with 3D resource models, 2D prospectivity maps, and grade-coloured drillhole intercepts — the same artifacts your geologists and investors expect, generated automatically from your raw site data.
Mining is the last industry to get its software moment. Lithium, copper, cobalt, nickel — the world needs more of all of them, faster. Today, joint inversion lives inside Python libraries only PhDs can run. Barq makes it a platform anyone can use.
If you're a miner, an investor, or a partner who needs to move faster on critical minerals — we should talk.