Insights

Mid-way through a roughly twenty-month applied-AI engagement with a major operator in Oman, the model was not the problem. Well deliveries had slowed enough to threaten the next phase, which needed a minimum of 10-15 wells to start in November 2022. Instead of escalating by email, the team ran a 39-slide interactive workshop that renamed one vague complaint - data collection is slow - as three separate, separately-owned problems: I can't access the data, I don't trust the data in the report, it takes too long to get to it. This is a template for any vendor blocked on the client's data supply chain: name the problems, show the honest infrastructure gap, and ask for a specific resourcing decision.

The middle of a project is where the case studies go quiet. This is the part they skip: Phase 2's descent into data-quality chaos on a major Oman operator's borehole image logs - a corrupt 0-15 pixel range, depth indices that disagreed with the log container, five wells rejected for missing apparent dip - and the climb back out through a dropped unsupervised detour to a first working supervised transformer, told through the dated running-log entries and the well-status ledger.

On a roughly year-long applied-AI engagement with a major operator in Oman, the artifact we trusted most was not the monthly steering deck. It was a single running-log deck that grew by appending every week: 288 extracted text blocks holding raw ground-truth-versus-prediction tables, a NaN-imputation status ledger that named its own failures (GAN and GAIN Fail, the masked autoencoder still Optimizing, KNN Succeeded with spikes), and the all-caps escalation the day the model hit a wall on missing data. This is an operating-model piece about why that living log, not the polished summary, is the real system of record, and how it later became raw material for two journal papers.

Reverse-time migration and full-waveform inversion are among the largest commercial high-performance-computing workloads on Earth. An exploration FWI project routinely consumes millions of core-hours, and the cost concentrates in the forward-and-adjoint wave solve: two anisotropic propagations per shot per iteration, times thousands of shots, times tens of iterations, on grids of billions of cells. For thirty years the response was to buy more cluster. The response now is to change the arithmetic of the solve itself. Neural-operator surrogates make a forward call up to three orders of magnitude cheaper. Differentiable simulators return the exact gradient by automatic differentiation, retiring the hand-coded adjoint. Mixed-precision computation and quantized inference fit bigger models on fewer cards. Compute-aware orchestration stops paying for solves that do not move the answer, and sample-efficient methods run fewer of them in the first place. Stacked, these levers do not shave a percentage; they change the regime: velocity building shifts from an overnight batch on a fixed cluster toward an interactive, elastically sized loop, and a solve cheap enough to run a thousand times turns a single best-fit image into a calibrated posterior. The limits are real: surrogates extrapolate silently outside their training prior, learned gradients are approximate, and on a real cluster the bottleneck is frequently I/O rather than FLOPs. The discipline the stack demands is the one it rewards: gate the surrogate, keep an exact reference in the loop, and measure the whole pipeline.

Why the well count and its intake gate were the real project-management artifact of a three-phase carbonate borehole-imaging program in Oman. The running tally read 5-in / 0-kept, then 10-in / 8-kept, then 25-asked / 8-delivered, and we tracked the asked-versus-delivered gap every month on a data-management dashboard. This is a first-person account of the ledger discipline itself: how we ran the program on one number so that, by the final report, no one was surprised by the 8.

The slowest, most expensive part of a carbon-storage programme is not injecting the CO2, it is deciding where to even look. An agentic screening pipeline gives each of the five geological suitability criteria its own software agent, reconciles them into one auditable GO / REVIEW / NO-GO verdict, and compresses a first-pass screening cycle from roughly four months per site to under ten minutes in a labeled synthetic demonstration, so a whole licence portfolio can be triaged in an afternoon.