It all starts
with how data is
read and written.
We do something you already do, but better. Same operations. Same data. More value out of every one. The reads-and-writes layer is what sits underneath the box, the hardware, V.L. and every chain we ship. Get this layer right and the rest builds itself.
Ask more. Store more. Retrieve faster.
Three things change about the data layer. They ride together — you don't get one without the other two.
Queries today's stack structurally can't
Per-tag downsampled SCADA tables can answer "what was the value at time T?" The reads-and-writes layer answers "what was the joint distribution of every channel at the same crank angle across every pump on the pad three quarters ago?" — in one read.
Lossless. A fraction of the footprint.
Frequency-ranked storage holds the full record set at much less disk than a conventional time-series database needs for the same data. Nothing thrown away. Nothing downsampled. Replay any moment from any point in the asset's life.
The answer is already in the chain
The storage layer is the audit layer is the feature store. One read returns a signed answer with the source rows attached. No separate audit pipeline. No separate feature pipeline. One read, one signed answer, one chain.
Three properties, by construction.
Not bolted on. Not patched in. These are inherent properties of how the data is stored.
Signed at write time
Every record is signed at the moment it's written. Tamper-evident. Replayable years later, identical to the day it was captured. Court-defensible by default, regulator-grade when the call comes.
Verifiable forget
When the substrate forgets something (per regulatory request or per customer instruction), the forgetting itself is a signed record. You can prove what was removed.
Honest unknowns
If the data isn't there, the answer says so. No fabricated values. No model-confidence smoothing the gap. The unknown stays unknown until evidence arrives.
Storage. Watchtower. Transport.
The foundation has three jobs. Most monitoring stacks split them across three vendors. We deliver them as one layer because they share one chain.
The chain that holds the asset's life
Every signal an asset emits, stored as it happened, signed as it's written. The chain IS the record of the asset.
Drift, peers, anomalies, evidence
The same storage layer answers the watchtower questions: who changed what, when did this start to drift, what does this look like compared to its peers, is this asset behaving like itself today.
How signals move around the pad
License-free radio between hubs and tablets. Cellular fallback. Tamper-evident at every hop. Lifts the recurring cellular line item at remote sites entirely.
The foundation decides everything downstream.
Every monitoring vendor in oilfield has chosen the wrong foundation. Their stacks are built on downsampled SCADA rows in a time-series database, with a cloud-trained model on top. That choice makes most of the questions a crew actually has structurally unanswerable.
- Pump-by-pump comparison at the same crank angle? Not possible — the crank angle was downsampled away on the way to SCADA.
- What did this bore look like at the same depth three quarters ago? Not possible — the database holds the last 90 days at full fidelity, then thins.
- Court-defensible record of what the asset actually did on the day of the incident? Not possible — the vendor's cloud-trained model is the only "evidence" and it's a confidence score.
- Verifiable forget when a customer demands deletion? Not possible — the model has the data baked in.
All four become routine the moment the foundation is right. That is why we lead with the foundation, and not with the application.
Compute, hardware, V.L. — in that order.
Once the foundation is in place, the rest is the natural consequence:
Compute — the VAM
A sealed box that runs the foundation on the asset, the same shape as the ECM on your engine.
Hardware — the library
Sensor packs, instrumented Rock Bits, ground scanners. 34 designs across 7 clusters, free to take with the license.
Application — V.L.
Validated Learning — the application that emerges when the foundation, compute, and hardware are in place.