In the small Australian town of Young in New South Wales, which looks to be about three hours from the fringes of the Australian Outback, Roman Teslyuk is getting ready for a prospecting expedition.
The chief executive officer of a four-year-old company called Earth AI, Teslyuk, who previously worked in mining exploration in Central Asia and the Middle East, is trying to reduce the environmental impact of one of the world’s dirtiest businesses.
The company uses a combination of machine learning and proprietary hardware to make the project of prospecting a bit more sustainable by improving the ways in which mining companies target potential sites to excavate and by reducing the footprint of the excavation work.
Mining, Teslyuk said over a Zoom call, is a lot like investing in a startup. There’s a thesis about what will work and the kinds of returns a project will produce, but the work is still incredibly risky with a lot of effort that can yield very little returns.
Earth AI’s proprietary prospecting vehicle. Image Credit: Earth AI
Teslyuk and Earth AI are part of a new breed of startups that are all looking to make mining more efficient by improving the process of identifying locations where companies should dig.
These are businesses like KoBold Metals, a collection of badass data scientists and engineers which use machine learning to pick sites to acquire and develop. Or LunaSonde, which bills itself as an “MRI for planet Earth”, and uses satellite imaging technology to find targets to explore.
Earth AI takes things one step further with technology that actually reduces not just the number of sites to excavate, but also the process by which that excavation is done.
“In order to feed the energy transition of the world — the resource industry will need to step-up. Roman and his team are well placed to lead this transition,” Balhuizen said in a statement.
Balhuizen is referring to the voracious appetite renewable energy generation technologies have for specific metals and minerals. Solar panels, lithium batteries, wind turbines, water electrolyzers, and electric motors require lithium, copper, cobalt, nickel, platinum, tin, and a whole laundry list of rare earth metals.
Earth AI attempts to solve that problem.
“We focus on finding new deposits,” Teslyuk said. Building a mine is an engineering problem, but finding where to mine is both a software and hardware challenge that’s a bit more technically complex to do sustainably.
“You can make a lot of waste if you go and drill everything out without considering how much you lose,” Teslyuk said. “We decided that… it was a no-brainer for us that we need to be more efficient.”
The company, which graduated from Y Combinator in 2019, first garnered investor attention for its use in the discovery of a Vanadium deposit — used to make large batteries for industrial applications.
Drill core from a Vanadium mining project. Image Credit: Earth AI
At the time the company claimed a 50x better success rate than traditional exploration methods at costs which averaged out to $11,000 per prospect discovery. That’s far lower than the millions that mining companies can spend on developing out sites.
The secret sauce is a system that uses drones equipped with magnetic sensors, a proprietary vehicle system that can carry testing equipment, and a waste management technology that leaves minimal disturbance on the surface of a site.
“Everything fits in a car and trailer,” Teslyuk said.
The lack of expensive drilling operations and outsourcing to service providers is what makes the project so compelling. “If you want to have a drill rig… that’s like $5.5 million,” said Teslyuk. “You have to get a bulldozer into the middle of nowhere and you have to make the tracks and make the roads and make the pads — and that’s for every drill hole. And then for every drill hole you have to make a ten-by-twenty meter pool that’s three feet deep for the mud system.”
To prospect in one area requires taking those steps for several different potential sites. “You have to give them a campaign,” Teslyuk said of the contractors.
“It’s inefficient. It’s unnecessary waste,” he said. “The pool is there to hold the mud and then you have to buy another bulldozer to fill the hole — and that’s just for testing.”
So Teslyuk and his team re-engineered the drilling hardware to solve the logistics problem. They built a closed loop, above-ground mud system that doesn’t require any digging. A series of tanks and pumps collect the mud from the drill hole and store the waste.
Then, when the project is complete and Earth AI has retrieved the samples it needs, the company simply fills the hole back up.
“When we leave it’s exactly like it was before,” said Teslyuk. “We built this integrated set of equipment on twenty feet containers or trailers and it takes us two days to move in and move out… That high mobility means we can change our program whenever we want… We can rethink our plan and be more efficient and cause so much less waste.”