The drill knows things. The question is whether you can hear them fast enough to act.
Every meter a drill bit travels underground, it generates signals—vibrations, torque readings, weight-on-bit measurements, rate of penetration changes, sensor data streaming up the drill string in real time. Embedded in that noise is geological information. What the rock is made of. Where the formation changes. Where the ore is, and where it isn't. Where the ground is about to do something unexpected.
Traditionally, making sense of that information required time. A lot of it. Raw downhole data would be logged, exported, handed to a geologist, and manually interpreted—often hours after the drilling event that generated it. By the time the analysis came back, the drill had moved on. The window to act on that insight had closed.
Propel built a system that closes the gap between signal and understanding to near zero.
The Interpretation Problem
Underground geology is not uniform, and it doesn't announce itself in advance. A drill passing through a sequence of rock formations encounters transitions that matter enormously—shifts from waste rock to ore, changes in lithology that affect bit performance, fault zones that signal structural risk, alteration halos that indicate proximity to a target. Each of these transitions carries operational implications. Speed up or slow down. Adjust steering. Flag for the geologist. Hold position and assess.
But those decisions are only as good as the information feeding them. And when that information is hours old, it's not really information anymore. It's history.
Manual interpretation workflows weren't designed for the speed that modern mining operations need. Geologists are experts, but they're working with logs that represent what happened underground, not what's happening now. The latency is structural—built into the process itself. And no amount of expertise fully compensates for working with stale data in a dynamic environment.
What Propel Built
Propel partnered with the client to build a fully integrated, real-time subsurface interpretation system—one that turns raw downhole signals into geological clarity as the drilling happens, not hours after it's done.
AI-Driven Geological Classification
At the core of the system are machine learning models trained to recognize geological signatures in streaming sensor data. As signals come up the drill string, the models classify what the bit is encountering in real time—identifying lithology, flagging transitions, and updating the subsurface picture continuously. What a geologist might spend an hour interpreting manually, the system processes in seconds. And it does it consistently, run after run, without fatigue or variability.
Real-Time Streaming Analytics
The system ingests data directly from downhole sensors and processes it as a live stream rather than a batch export. That architectural choice matters more than it might seem. Batch processing introduces latency by design—you wait for the batch to complete before analysis begins. Streaming means the analysis is always current, always running, always one step behind the drill rather than several hours behind it.
Intuitive Dashboards
Raw classifications aren't useful unless they're visible. Propel built dashboards that present the interpreted geology in a format operators and geologists can actually use in the field—geometry, lithology, and risk zones rendered clearly, updated in real time, and designed to support fast decisions rather than extended analysis. The interface puts the right information in front of the right people without requiring them to dig through data to find it.
Automated Alerts for Geological Transitions
When the system detects a major geological transition—an approaching fault zone, a shift from waste to ore, an anomaly in the sensor signature—it doesn't wait for someone to notice. It sends an alert. Automatically. Immediately. So the team can respond to what's happening underground right now, not to what happened while they were looking at something else.
The Results
Interpretation time collapsed from hours to minutes. That's the headline number, but the implications run deeper than a simple efficiency gain. When interpretation is fast, decisions become proactive instead of reactive. Teams adjust in real time rather than after the fact. Deviations get caught before they compound. Ore contacts get confirmed while the drill is still in position to act on them.
Accuracy and consistency improved too. Machine learning models don't have bad days. They don't interpret a log differently on a Friday afternoon than they do on a Monday morning. The geological picture they build is grounded in the same criteria every time, which means it's also comparable across runs, across sites, and across time—something manually interpreted logs rarely are.
In-field decision-making got sharper. Operators had better information, delivered faster, in a format they could actually use. The result was fewer hesitation moments, fewer calls back to the office to wait for expert review, and more confident execution at the drill face.
The Foundation This Creates
Real-time subsurface intelligence isn't just operationally useful. It's structurally necessary for what comes next.
Autonomous drilling systems—drills that navigate, adapt, and make decisions without continuous human input—require a continuous, accurate picture of what's underground. They can't wait for batch-processed logs. They can't operate on stale geological classifications. They need the same kind of real-time awareness that a human expert would bring, delivered at machine speed and without the latency that human workflows introduce.
What Propel built here is that foundation. A system that sees what's underground, understands what it means, and communicates it fast enough to drive action. The drilling intelligence that comes next gets built on top of this.