Data access
Good marine data is licensed and hard to obtain. Without it the methods stay on synthetic tests.
Gate. Continue to the next phase only if at least one licensed or open dataset can be used under clear terms.
Investors
A thesis we can be wrong about, and the tests that would show it
This page states what we believe, how each belief will be tested and what could stop the program. It does not describe results, because none exist yet.
Four hypotheses
Each is written so that an experiment can prove it wrong.
H1 Better deep-marine imaging
ConceptPhysics-based joint inversion can improve deep-marine subsurface imaging over current industry baselines.
How we will test it. Run the method and the industry baselines on the same synthetic and open benchmarks, with the same inputs, and compare the images and their uncertainty. Report the cases where we do not improve.
H2 Uncertainty that is useful
ConceptReporting uncertainty with every image lets a decision-maker act with more confidence than an image alone.
How we will test it. Check on benchmarks with a known answer whether the stated uncertainty matches the actual error, and ask independent reviewers whether it changes their reading.
H3 Methods combine
ConceptSeismic, gravity and electromagnetic data constrain each other, so a joint model is more reliable than any one alone.
How we will test it. Compare joint and single-method inversions on a model where the truth is known, and record where combining does not help.
H4 The cost can be managed
ConceptThe compute needed is within reach of a small team using modern accelerators and careful engineering.
How we will test it. Measure the compute used by each benchmark run and publish it with the result, so the cost is a recorded fact and not a guess.
Where the program is going
The phases are the same as on the Research page. Only the current one is under way.
- P0FramingCurrent phaseDefine the questions, the claims registry, the evaluation rules and the limits.
- P1Synthetic benchmarksPlannedBuild synthetic models with known truth and fixed baselines.
- P2Solver validationPlannedCheck the wave solver against analytic cases and published references.
- P3Open-data inversionPlannedInvert openly licensed datasets and compare with published results.
- P4Uncertainty testsPlannedTest whether the stated uncertainty ranges contain the truth.
- P5Joint inversionPlannedCombine data types in controlled experiments.
- P6Independent reviewPlannedExternal reviewers examine the methods, the data handling and the claims.
- P7Pilot readinessPlannedPrepare a pilot with a partner, on data they are licensed to share.
What could stop us
Six risks, each with the gate that decides whether the program goes on.
Compute cost
Joint inversion is expensive. The cost may not fit a small team.
Gate. Stop scaling up if the recorded cost per benchmark rises faster than the gain in the result.
Scientific validity
The central hypothesis may be wrong. The methods may not beat the baselines.
Gate. An independent review of the benchmark results decides, and a negative result is published as it is.
Talent
The program needs rare skills in geophysics, numerical methods and engineering.
Gate. Do not start a phase without the named expertise to run and review it.
Dual use and export
Subsurface imaging can be misused, and some methods may be subject to export rules.
Gate. Civilian applications only. Legal and export review before any release of methods or data.
Market timing
Demand for this kind of imaging can change faster than research can.
Gate. Review the direction at the end of each phase against what partners actually ask for.
Request a briefing
Tell us who you are and what you would like to understand. A person reads every message.