Given the same data, Mohan builds a better subsurface model than any existing tool.
Mohan builds generative subsurface world models for porphyry copper exploration. We combine real drillhole data, physics-constrained modeling, and expert geological reasoning to produce probabilistic 3D subsurface models.
Subsurface uncertainty, unaddressed
Industry-standard tools produce a single deterministic model based on intuition. One interpretation. No measure of confidence.
You get one model, not a probability distribution. You cannot know how confident to be in any given interpretation.
Interpolation does not enforce geological laws. The resulting model may be statistically smooth but geologically implausible.
The geologist's intuition shapes the parameters, but this reasoning is never made explicit or reproducible.
A generative subsurface world model
Mohan fuses three data sources to produce not one model, but a probability distribution: tens of thousands of geologically valid subsurface architectures consistent with your data.
Treated as hard constraints that any valid model must honor.
A generative model trained on synthetic porphyry copper systems consistent with geological formation laws, providing the best possible prior over subsurface architecture.
Geological knowledge encoded as soft constraints via LLM reasoning, integrating how expert geologists interpret data rather than just what they conclude.
An integrated three-part architecture
A synthetic data generator, a diffusion-based generative model, and a drill sequence optimizer, all working together end to end.
Value across the mine lifecycle
From early-stage exploration through production, Mohan fits into your existing workflow.
Optimize drill sequencing, reduce total holes needed, and shorten time to resource definition.
Guide mine development decisions with calibrated confidence intervals, uncertainty maps, and better ore body delineation.
Better models, better decisions
Quantify uncertainty at every point, optimize drilling sequences, and make mine development decisions backed by data instead of intuition.
Not a single best guess, but tens of thousands of geologically valid models consistent with available data.
Generate model samples, score candidate locations, and prioritize where to learn the most.
Use calibrated confidence intervals and uncertainty maps for downstream planning.
The team behind Mohan
Geologists, AI researchers, and engineers from MIT, building together.
Company strategy and operations, investor relations, VLM Shadowing pipeline architecture, expert reasoning capture, and commercial partnerships.
Director of MIT Earth Resources Laboratory. Scientific credibility, external validation, and warm introductions to Tier 1 client targets through extensive relationships with junior and mid-tier exploration companies.
Physics-constrained synthetic data generator design and validation, geological domain authority for model architecture, and client-facing geological credibility.
Diffusion model architecture design and training, physics-constrained generative model development, and core ML research execution.
Product positioning and design for go-to-market, client-facing product narratives and interfaces, and product experience from first client contact through delivery.
Engineering execution and technical infrastructure underlying Mohan's modeling and analysis platform.
Principal Research Scientist at MIT Media Lab and Director of the MIT Minerals Stewardship Consortium. Executive-level relationships at BHP, Vale, and Rio Tinto.
Professor at MIT DUSP and IDSS. Extensive experience advising deep tech companies on B2B go-to-market strategy in regulated industries with long procurement cycles.
Assistant Professor at MIT EECS and MIT Media Lab. Research on multimodal foundation models and human-centered AI systems. Multimodal architecture for VLM reasoning capture pipeline.
Research Scientist at MIT ERL. Former holder of the Lundin Chair in Economic Geology at University of Arizona. Direct relationships with geologists at Freeport-McMoRan and Lundin Mining.
CEO of Butlr Technologies, Forbes 30 Under 30. Built Butlr from MIT Media Lab research to a funded company with enterprise clients. Commercialization strategy and fundraising advice.
PhD Candidate at MIT IDSS. Co-author of HugAgent. Research on LLM reasoning architecture for expert knowledge capture and VLM Shadowing pipeline design.
Continue building the product surfaces
The structure now follows the company document more closely. The next step is to replace visual panels with actual product visuals, geological outputs, benchmark figures, and system diagrams.