Recently our VP of Business Development, Sahiba Sachdeva, sat down with Martin Veitch of Mining Journal to discuss how Descartes Labs' technology is empowering mining companies to make sense of incredibly complex datasets with interfaces geared towards specific problems including mineral exploration and infrastructure monitoring.
Watch the Video
Transcript:
Martin: Hi everybody and welcome. This is Martin Veitch for Mining Magazine and Mining Journal once again. With me, I've got a very special guest in the studio today, Sahiba Sachdeva. Sahiba is the VP of Business Development at Descartes Labs. We're going to be talking about all things geospatial, remote sensing, data science, and some of the most cutting-edge technologies, as they relate to modern mining.
Sahiba, welcome.
Sahiba: Hi, Martin. Thanks for having me.
Martin: I understand you're joining us all the way from very glamorous Washington, D. C.
Sahiba: Yeah. I don't know about glamorous, but it certainly is a busy place these days.
Martin: Why don't you tell us a little bit about the company, its history and where you are today in the mining sector?
Sahiba: Yeah, absolutely. So, Descartes Labs was founded by a group of scientists at coming from the Los Alamos National Labs in New Mexico. And folks might be familiar with Los Alamos from the movie Oppenheimer. Although that, wasn't really what they were focused on. Essentially, our founders observed that there was a growing number of sensors for remote monitoring of this earth being launched all the time.
There wasn't adequate technology infrastructure to really work with that data effectively, much less get the value that was promised from it. So, they founded our company to establish that technology. Our engineering team has continued to refine upon it and build upon the technology over time.
And we have the world's 40th most powerful supercomputer available at our disposal today. And that trend that they observed has really just continued. There are more and more satellites launching all the time. You have different modalities of data. You have airborne collections happening. The complexity of the data and metadata continues to grow.
Spectral resolution of data continues to become finer. And so, working with that data continues to be just very difficult to do. We're talking about really big data over here. Actually, a client of ours was joking yesterday that, people think they understand how big this data is, but they don't, it's kind of like when you think, Oh, I can be a billionaire, but you don't really know what that means.
The size of these datasets are actually huge. And we have some clients who've spent millions of dollars on data sets. Sometimes, and they can't even work with them effectively until they bring them into our technology suite.
The technology that we have is very advanced, but what do you even use it for? What do you use remote sensing for? And we use it in a number of industries. Certainly defense, but also things like natural catastrophe response, agriculture and mining, which is what we're talking about today, of course, and in mining, we use it for a number of use cases.
Mineral exploration is a major one, so looking for areas to target during field campaigns and drilling campaigns. And then also operations monitoring, safe operations monitoring, and then finally rehabilitation and closuremonitoring.
Martin: Your organization has been in geospatial since I think about 10 years.
Is that correct?
Sahiba: Yep, that's right. So yeah, we were founded about 10 years ago and have been working at this ever since.
Martin: And tell me how widely deployed is this stuff? I mean, is it typically for the elite? As you talk about those data sets, accessing those, managing them, having the infrastructure to deal with those can be challenging for many organizations.
How widely deployed would you say geospatial has been?
Sahiba: So, it's interesting you asked that, and I would say that the, level of use really varies by the use case. In mineral exploration for example, we do see a fairly large community of users who are advocates for using remote sensing.
But even so, there's been folks over the years who have over promise what remote sensing can do. There are some parties who have been disillusioned. And so sometimes when we are talking to folks about this technology, we do have to sort of educate on what is really possible, which is plenty.
But that is sometimes something that we have to get after. Our company has always been very firm on adhering to scientific ethos and principles, but we never would, mislead or over promise on what's really possible with remote sensing. And so, we do have a really strong reputation in this industry.
A lot of trust with our clients, which is fantastic. That's something that we have really earned over the 10 years that we've been doing this. In other use cases like operations monitoring, I think that it is still pretty, there's plenty more that can be done there. I think we're sort of early in that.
There are some use cases that are fairly established, like monitoring of tailings dams using in SAR data. I would say that we consider remote sensing to be complimentary to monitoring that's already happening on the ground. We wouldn't say that it's a replacement, but it does provide a lot more context and a lot more of a holistic view of assets than what can be achieved by point solutions. For example, with a tailings dam, you will have individual sensors placed on the ground along the wall, remote monitoring is almost like blanketing the entire dam wall in sensors. So, it adds a lot more context to what you can do on the ground.
I would say for rehabilitation and closure. It's really early. I think a lot of metrics, are around things like tree counts and, volume vegetation that the metrics are geared towards physical manual, measurement and not around remote measurement.
So, I think everyone understands that remote measurement is certainly the way to go for areas that are difficult to access. We don't get as much foot traffic as they did in their operational heyday. But we have a way to go to work with regulators in terms of figuring out what metrics are really going to be meaningful for those use cases.
Martin: Yeah, absolutely. But to go back a little, you talked about the sheer scale of these data sets. For those organizations that don't own or don't have access even to these very large data sets, or they don't have the engineering know how, the programming skills, the data science folks who are incredibly in demand and thin on the ground at the moment, what's the situation for these guys as a way to bridge the gap?
Sahiba: Yeah, absolutely. So, a number of the data sets that are used in, these use cases are actually free. They're being collected by governments, by the European Space Agency or NASA or other agencies. The German and Italian space agencies have hyperspectral data sets that are, free for use today. So, there is a lower barrier to entry from a cost perspective.
Some of these use cases still, from these agencies, you can get the data in raw form, but the know how around how to process it and make it analysis ready, how to stitch together scenes that are side by side and then how to actually get insights from that data is fairly niche.
There are folks who study remote sensing and know how to do this. But it is definitely a specialized field. What we really focus on is making that as accessible as possible. All of the data available in our catalog is going to be already preprocessed and made analysis ready for end users, or we will have put them into end user applications where analysis of that data is sort of more of a point and click experience versus a data science experience. Now, if you are a data scientist, there's flexibility to be able to work with that data, as creatively as you'd like, but we do try to lower the barrier of entry to other folks.
Martin: Yeah, and I know also while we're talking about these incredibly complex technological phenomena, tell me a little bit about where we are with some of these kinds of techniques today and where they're going to go next.
Sahiba: Yeah, absolutely. For most of mineral exploration, remote sensing history, folks have used multispectral data sets. Aster has been the workhorse of remote sensing for mineral exploration for a number of years, and it's still very much in use today.
The thing is that with multispectral datasets, you can at best distinguish mineral groups, but not mineral species because, you're looking at fairly, broad spectral ranges when you are looking at the earth. Now with hyperspectral dataset, you have access to hundreds of spectral bands.
It gives you a lot more granularity in terms of what you're looking at. You are now able to distinguish mineral species as opposed to, mineral groups. And we have a blog post on our website, sort of a case study over Cuprite . If you look at it with the multispectral data, It's hard to tell whether what you're looking at is alunite or dickite or kaolinite, but then you look at it with the hyperspectral data and you're able to start to distinguish those, which gives a lot more detail to geologists when they're looking at a site.
The availability of hyperspectral data is essentially giving us the chance to relook at the entire world all over again and learn much more about each site that we previously thought we knew and understood.
Martin: Tell me a little bit about the products that you have today. I know charmingly named products, Iris and Marigold, always feature very highly here.
What's the product development plan and what's coming next in those product lines?
Sahiba: Yeah, so Marigold is our product for mineral exploration and uses remote sensing, multispectral and hyperspectral data sets. The hyperspectral capabilities that we have in there are fairly new and the thing that we're very excited about and, are continuing to refine for our client base today.
We built the hyperspectral processing capabilities into Marigold really when our clients made a very strong case for us to do so. They sort of challenged us to do so and they said, you know, we've looked around at the market, looked at the industry, and we're very excited. We don't think anybody else can do this if you guys can't do it.
So, you really have to do it. We rose to that challenge. And I think everyone's very excited about what is possible now in Marigold with the hyperspectral capabilities. And so that's really going to be our focus for the near term is continuing to develop and refine what we can do with the hyperspectral data.
Iris is our product for operation safety monitoring. Right now, it is focused on monitoring tailings dams and monitoring motion and deformation on tailings dams. Of course, it can be deployed towards other assets as well. Iris is actually fairly new. We, came out with this user interface to be able to interact with the ground deformation data just this year.
And we are working with our clients on refining the user experience and, starting to bring in additional metrics that would complement their understanding of the tailings assets. Whether that's, you know, pond monitoring or vegetation on the dam wall, we're working with them to understand.
What will be the most relevant and helpful metrics for them to monitor and bringing those into the interface.
Martin: Yeah, tell me about the philosophical differences in what we do with all this data and these incredible, the incredible scale of these assets we have at our disposal now, opposing that ethos of data is king and all of these cliches and replacing it with practical actionable insights that can be derived from data. Tell me a little bit about the difference in those two philosophies as you see them.
Sahiba: Yeah it's just that we've found that it's very difficult to get insights from that data unless you really know what you're doing.
So we do try to provide as much of an on-rails experience, as much guidance and training and specific workflows for the use case to our clients so that they can make the most of that data, because if you don't know what you're doing, you can get results that could be misleading to you and then you waste your time and energy, right?
We really do try to provide much more of an on-rails experience than what you can get if you just had the data. The data is, of course, very powerful. And we're sort of spoilt for choice. These days, there's more and more coming online all the time. It's just not enough without the analytics to complement that data.
Martin: Yeah, there's always that question, isn't there, of garbage in, garbage out, and not seeing the wood for the trees, as with anything with data management. Look, we'll have to wrap up in a second, but you know the situation in mining at the moment. It still gets a lot of criticism for erstwhile practices, but at the same time, it's becoming embedded and suffused with digitization and automation tools.
Are you confident and optimistic about the future of mining generally, and how it, combines with the data world?
Sahiba: Absolutely. You know, I'm fairly new to the mining industry myself. I just started working in it about three and a half years ago. But I've loved it ever since I've joined. It's just an incredible community of people who are so smart and care so much about the work that they do and about the footprint that they have on this earth.
The mining industry is critical to how we get over our reliance on the oil and gas industry as a society. And I think that the industry does wear it with a lot of, responsibility. We're really happy to be part of that journey and be part of that solution.
But we do it because we see that our partners in the industry carry that same responsibility with as much care as you'd like to see, and there's so much innovation. When you go to these conferences and you meet with folks in the industry, there's so much excitement and so many good ideas for how we can do this effectively, and we're just really happy to be part of that solution.
Martin: Well, it's a vast topic we could be talking about Sahiba, but you've done a great job, outlining some of the core trends, and I'm sure giving a lot of our viewers food for thought. So, Sahiba Sachdeva, I want to thank you very much for your time today.
Sahiba: Thank you, Martin.