Satellite Data and Conservation: Gracie Ermi on Land Use, AI, and Protecting the Planet

Gracie Ermi, Machine Learning Scientist at Impact Observatory, highlights how AI and satellite data transform land use mapping and environmental protection. By analyzing satellite imagery with machine learning, she improves data analysis efficiency and accuracy. Gracie discusses AI's impact on wildlife conservation, interdisciplinary challenges, and the importance of trust with environmental experts. She also shares her career journey into satellite data and encourages women to explore computer science and AI in conservation.

Takeaways:
🤯 AI is enabling the geospatial field and environmental protection to expand in ways that weren't possible before.
🌍 Creating land use maps once took years. With AI we can produce global maps in weeks!
🌲 Find datasets and play with them to get involved with conservation and AI!

Resources Mentioned:
🌟 wildlife.ai
🌟 fruitpunch.ai
🌟 opensustain.tech
🌟 Conservation Tech Directory
🌟 AI for Conservation Slack
🌟 Wild Labs
🌟 Zooniverse

Podcast:
🗳️ Season 1 Feedback? Questions?
🗳️ Nominate Podcast Guests!

About Gracie:
Gracie Ermi is a machine learning scientist specializing in building machine learning & AI applications to support wildlife research and inform environmental policy. She uses aerial imagery, remote sensing, bioacoustics and more to study marine and terrestrial species, and monitor global land use and land cover. As one of the co-creators of the Conservation Tech Directory, she works to encourage more connection and communication across the conservation technology field. Gracie also prioritizes educating and inspiring the next generation of innovators, and as an AAAS IF/THEN Ambassador and an AGU Voices for Science Fellow, she has spoken to over 150 audiences around the world about the intersection of conservation and technology.

Connect w/ Grac

Chapters

• 0:00 Introduction and Background
• 3:01 Mapping Land Use and Land Cover
• 5:25 Data Sources and Updates
• 6:37 Use Cases and Applications
• 7:39 Frequency of Imagery and Real-Time Data
• 9:06 Satellite Data Sources
• 9:37 Role of Machine Learning in Geospatial Analysis
• 10:40 Labeling and Categorization of Land Features
• 11:39 Improving Labeling and Data Quality
• 12:40 Surprises and Innovations in the Field
• 14:25 Data Size and Storage
• 14:57 Observing Climate Change
• 17:23 Geospatial Analysis vs. Computer Vision
• 17:46 The Future of Geospatial Analysis
• 19:58 Unlocking Hidden Insights in Vast Datasets
• 21:41 Evolution of the Field
• 23:10 Technology Adoption and Anxiety
• 24:26 Gracie's Career Journey
• 26:26 Women in Computer Science
• 29:03 Further Education
• 30:36 The Future of Computer Science
• 32:25 Getting into the Field
• 34:05 Gracie's Recommendations and Resources
• 37:27 AI Conservation Volunteering and Data Sets
• 39:25 Podcast Guest Nomination Form

Transcript

0:00

Hey there, as we wrap up season one, we are thrilled to share with you that season two is already in the works, woohoo! We've started recording with some fabulous guests and stories I'm confident you're gonna love. But my top priority is serving you, our amazing listeners, and we would love your feedback. What questions are on your mind? What intriguing things do you wanna make sure we cover in season two? Please help us make season two even more epic together, please head to our form in our show notes. You can find it at yourairoadmap .com / podcastfeedback. Share your thoughts, give constructive feedback. You might even be featured in a shout out of a future episode. Thank you so much, Tim, for that awesome idea, et cetera. We are so excited to hear from you and grow and make this podcast even better. So excited to hear from you. We will read every single one, I promise. Okay, back to the show! Hi, my name is Joan Palmiter Bajorek. I'm on a mission to decrease fluffy hype and talk about the people actually building in AI. Anyone can build in AI, including you. Whether you're terrified or excited, there's been no better time than today to dive in. Now is the time to be curious and future -proof your career, and ultimately, your income. This podcast isn't about white dudes patting themselves on the back. This is about you and me. and all the paths into cool projects around the world. So what's next on Your AI Roadmap? Let's figure it out together. You ready? This is Your AI Roadmap, the podcast. Hey folks, this is Joan. I wanna give a little background about this episode with Gracie. Gracie and I spoke at Geek Girl Con a few years ago. We were on the same panel and I was like, she's so cool and doing such amazing work. And I followed up with her because I don't know about you, but I don't think that much about geospatial stuff. Satellites are not necessarily part of my daily work. There's a new project, so who knows? Maybe that will happen, but. Gracie is dealing with this type of data all day, every day. She thinks about this from a computer science and machine learning perspective. I am so delighted to learn from her. I think in this one I asked some really weird questions about animals anyway. I'm learning too. I'm learning with you, expanding my knowledge set of different parts of the field. Gracie has such a cool perspective about how her education was shaped by other people who really wanted more women in AI, in machine learning, in computer science. It's so powerful. I think you're gonna learn so much. I had a delightful time speaking with Gracie and I hope you have a delightful time listening. All right, let's jump in. Hello, hello. Great to have you here. Would you mind introducing yourself? Yeah, my name is Gracie Ermi. I'm a machine learning scientist at Impact Observatory. Okay, so can I ask what are you working on? Yeah, so at Impact Observatory, we build global maps of land use and land cover using AI and satellite data. So maps show like where are crops, where are there trees, sort of different classifications of land. And the earth is rapidly changing. So, you know, new cities are built, trees are taken down, like crops are planted or they're harvested. So we're trying to map how is that land changing as time goes on. and we can build these maps over any time period and any area that a user might be interested in. And so what I've been excited about recently is kind of how do we use these maps and then go one step further and pull out insights about where the most relevant change has happened, where's the most important change that's happened or like unexpected potentially change that has happened. How do we automatically detect that and show it to the people who are using these maps, who tend to be people who are working to protect the planet. They're trying to protect areas of land that need to be protected or trying to monitor how their projects are impacting the land as they're building new things. So there's a lot of really important decisions and things that can be made using these maps as. a data point in those decisions that get made on behalf of the environment. So cool, well, I have so many follow-up questions. Okay, so in my mind, and you're welcome to tell me this is wildly inaccurate, but my brain is mapping it as kind of like a Google Maps like overview, maybe with higher fidelity or kind of, is it grayscale? Like when you're trying to describe it to someone who hasn't seen it, what does it look like? Yeah, yeah, that's a really good sort of frame. So like Google Maps plus some extra data basically, or kind of, and also more frequently updated than Google Maps in terms of what the actual land type is. So Google Maps will show you, there's roads here and here's where your grocery store is kind of stuff. And then our maps are. They're really colorful if you look at them. So it's like different colors mean different types of land. So in our maps, like red is like built area. So neighborhoods and cities and things, orange is crops, green is trees. So if you look at like our map over the whole state of Washington, it's like red around Seattle. There's a lot of green out on the peninsula and like a lot of crops, a lot of orange crops on the east side of the state. So it shows you kind of a bird's eye view of like, what does the land look like? in this time period when we ran the maps and as time goes on we can update them as the land is changing and see those changes. That's so cool. When are you getting this data from satellites or internal data sets or how do you update Yeah, it's satellite data. So we're using a few different types of satellite providers. One of our main resources is Sentinel-2 data, which is all freely available for anyone to use, which I think is really cool. I really like that part. And some of the maps that we produce with that data are also openly available. Our annual maps of the entire globe are openly available, so anyone can go. look at them and download them and use them, which is also something I'm really excited about. And we also use some higher resolution data for higher than Sentinel-2. So that's a little bit more expensive, but you can see more things because it's higher resolution. Totally. hmm, I'm processing, you're saying we have satellite image data of the whole planet, is that correct? Yeah. Okay. it's like less frequent at the poles. And this is also, I'm like new to the satellite data industry and world, there's so much there to learn. So I'm also learning this kind of stuff as we go. But yeah, we can get imagery of the whole globe. Okay, yeah, okay. I'm learning something new every day. Okay, so how might people use this strategically? You said people are using it for environmental things. Can you give us a few use cases or examples? Yeah, yeah, so if you're someone who manages a protected forest or something, you could order maps over your forest and see, are there trees everywhere? You expect there to be trees? Did something change? There's also, with natural disasters that have occurred recently, like with Hurricane Adalia that happened last year, it hit parts of Florida. we were able to generate maps of that area before the hurricane and then right after the hurricane. And they're actually used by search and rescue on the ground to like see where, where are their potential areas where people could be stranded that, just to make sure they checked all the possible areas. So there's use cases like that. If a company has a new construction project, They could use our maps to be able to track how the construction was going and whether maybe more trees got taken down than they expected or something. They could see that in the maps. And we can track sustainability and green promises that are being made by companies. They can use this data to make sure those promises are being kept. Yeah, that's so cool. One, how often, so I'm thinking about the hurricane example, which is sad, but really thinking about how it can be leveraged for safety or otherwise. How often is it getting the imagery? Because I'm thinking about if I'm a stranded person, you know, in the hurricane, are they gonna like minute to minute know where I am on the log or like, or how often is that data being taken? Is that a silly question? No, that's a great question. Yeah, that's a really important distinction too. So it's not, it would never be exactly real time because the satellites, you know, like with the Sentinel-2 imagery, we get new images every like 10 days, I think, or so, seven to 10 days around there. And the higher resolution satellites are a little bit more frequent than that, but it's not like we can get the images like right after. necessarily some event happened. So the more important thing there, I think, was those, when we know that a disaster is going to hit soon, like a hurricane. Some disasters we, of course, can't plan for like that. But with a hurricane, we know that it's probably going to hit these areas. So generating those maps of the area beforehand can be really, it turned out, was helpful to those search and rescue missions, because they could go in and see where buildings were and make sure that they, you know, could see all the areas where people might be beforehand. And then, yeah, we got them the post hurricane map as soon as we could. But you're right, it wasn't like, yeah, you need like different technology, I guess, to be able to tell, you know, where people are like right after. But it was still, you know, another piece of data to add into that planning process. Well, and you're mentioning Sentinel-2, is that a government agency? Is that a private entity or where does that come from? Yeah, it's a European space agency that's their satellites. Very cool. When you are, okay, so you also mentioned that, so you're leveraging machine learning for these. How are you thinking about leveraging, or I feel like a lot of people are like, throw AI on it, throw machine learning on it. How could you just spell, like, how is it helpful? How do you use it? How do you think about machine learning in these contexts? Yeah, yeah. So there have been efforts ongoing to create these land use and land cover maps for a long time before this. And usually these maps come out like every two years at the most. And so you have to wait until two years have passed and then you get a map of the earth as it was two years ago. So what's cool about AI is like, so we have all these satellite images that are being taken at this, you know, rate. And we can, to make these maps, previously before AI was being used, someone would have to go in and label all the different points in all of these images. So like, okay, there's trees here, there's buildings here. And I'm sure they used some, maybe some automation within that pipeline. But with AI being added in, we can train it to look at an image and tell us what's... what's on the ground in that image. And so that just being able to run that for images of the entire globe is like, it's much faster with AI. So we can make these global maps in a matter of a few weeks rather than a few years. So it just really increases the, we're much closer to real time than without AI. Whoa, a few weeks instead of a few years? Is that what you just said? That is, wow, okay. Well, and I don't mean to ask anything too private, but what are you labeling it with? Just like, this is a tree, this is a pine tree, or like, what, is that again, a silly question? no, that's not silly at all. Yeah, our maps are the categories right now, like the main sort of categories are tree, crops, built area. We sort of have differentiated roads and buildings and water. And we're kind of trying to pull apart those more general labels as we go and. create more distinction between those different labels. But in the like, yeah, at the most basic level, we're not worried too much about labeling the types of trees or the types of crops. Yeah. Yeah. One is also, do you have like human in the loop to test the fidelity of the labels? Do you all have certain like, you know, in other fields there, you know, percent metric of error rates kind of thing? Do you all think about that way or how do you like, as much as you're willing to share like the process of these things? I'm so curious. Yeah, um. We're sort of working towards that. We definitely want to, well, as I said, the land is changing rapidly. So we don't want to necessarily use super old labels or anything. We want those to be sort of more up to date. So yeah, we're working on how to make better use of the expert labeled images that we have. And yeah, potentially. update them, improve them over time. And yeah, I'm not sure how much I can say about that. But yeah. in my mind, the parallel I'm making is about kind of content moderation in general and how, you know, different platforms, what they moderate, do they choose to moderate, you know, what my label is, you know, hate speech, all these different tags are done by humans in algorithms behind the scenes. And, you know, we noticed when Twitter fired a lot of people in that, you know, how that... radically changed the content that was being seen by users. Trees and buildings, totally different use case. But thinking about just the rapid speed you're talking about of efficiency is like, my brain is struggling to process. And when you, so how long have you been working on these type of projects? Yeah, I've been there almost two years now. Okay. Are there things that have surprised you along the last two years or innovations that you're like, whoa? Yes, I feel like, oh gosh, like the sort of geospatial world is so vast and like there's so much to learn there. So every day I feel like I'm learning something new or like finding out some new thing you can do. It's something that's been really surprising is like how much information you can pull out of a satellite image. Like there's, you know, it's not just red, green, blue. regular image. There's like these extra channels of data that where you can measure things like how is the vegetation doing in that image, like how healthy is it. You can find water in an image just from combining the bands in different ways. There's like all these extra analytics you can do with just a satellite image which is like really fascinating to me. And then on the other hand, a surprise has been like how much clouds can hinder your work. Because you can't control them and you can't see through them with just a regular satellite image. So, yeah, clouds are like my number one enemy now. Oh, enemy. Yeah. Well, and can you see like down to a blade of blade of grass? Like how close do you get down to the minutiae? Yeah, definitely not that specific. So the Sentinel-2 images that I keep talking about are 10 meters per pixel. So anything that's smaller than 10 meters across, you can't see it in very good detail. But if it's bigger than 10 meters, you can start to see it. And then some of the other imagery we're using, Planet Labs. has satellites that we use as well and they're three meters per pixel. So starting to see more detail, especially in like buildings and roads and things. But yeah, and then there are satellites out there that are higher resolution than that. But yeah, that's about where we're at. I'm imagining these must be quite large data sets, like... yeah, they are very large. Like, can you give a range of like gigabyte, terabyte? Like, how big are we talking? Yeah, I mean, so we luckily don't have to host the images ourselves. We use Microsoft hosts, at least the Sentinel-2 imagery, and so we can utilize a lot of their tools to manipulate it, work with it, pull it down when we need it. Yeah, I don't even know if I could give an estimate of how much data there is, but it is a lot. Yeah. Okay. This imagery data I know is very, very heavy. Um, that's, that's fascinating. Um, when, I mean, you're talking about kind of these benchmarks of two year periods and, you know, we're experiencing climate change, you know, pretty rapidly. Do you see that in the dataset? Is that something that's tangible? Is that kind of just things are changing? Like, do you, I guess, do you see that? I should just start there. Yeah, so we can definitely see some of the effects of climate change. Like we're kind of starting, well, the people who use our maps, who order maps from us, they kind of take the maps and potentially can pull out things that are related to climate change that are relevant to their use case. We're definitely starting to move towards how do we... generate those insights, like with the change detection, how do we automatically let someone highlight some change that could be related to climate change? And as we move forward, we're trying to pull out more and more useful insights. And so those will, I'd say inevitably be related to climate change because so many changes that are happening on the earth are related to climate change. And yeah, so it's... And... Environmental protection is a huge part of why we're building these maps and what we want to be able to move towards with the maps that we're building. But it's kind of another step of analytics to be able to say, oh, this was because of climate change. Yeah, no, that makes sense. I'm just curious as I hear benchmark these things. I guess this is a use case that I just haven't heard of until I met you. How big is the company? How, I don't know if you can share like how many customers you have. Like, I'm just very curious about, is this very common? Are there tons of you that I don't know what people working on this? Like, yeah. yeah, yeah. So there's less than 20 people at Impact Observatory. We're still kind, we're a startup and so we're still, you know, finding customers and figuring out kind of what are the most helpful products that we can build that are in this mapping land use and land cover, you know, world that we can sort of supply. And there are definitely other companies out there working on this type of stuff. Definitely, like you said, like more than I knew for sure that we're in this geospatial AI world. Like it's a, it's really fascinating to, you know, learn about all the different things you can do with satellite images and the things you can measure and learn and monitor. Yeah, well, and as you talk about kind of geospatial, I guess my mapping is kind of computer vision and image recognition. Are those really similar to geospatial? Is that a completely different, can you kind of describe the differences or similarities between those? Yeah, yeah, I would say geospatial is more like purely space data, like satellite data, and analyzing that, and then that can kind of intersect with, you know, computer vision and AI in various ways. So just because you work in geospatial doesn't mean you're necessarily using AI, but I think more and more AI is being introduced into that field. for good reason. I mean, you just told me things that blew my mind. So yeah. this is one of the bigger questions. So I think we hear a lot of people say like, in the next five years, blah, 10 years, all these extrapolations. And some people on this podcast have been laughing at me and been like, what is the next six months? When you think about like the biggest, broadest part of your field and these types of projects, where do you feel like your field is headed? Yeah, so I do think like AI is really enabling the geospatial field and the environmental protection, like wildlife conservation, like all those fields to expand in ways that they wouldn't be able to otherwise. So you know, we talked about like how much satellite data there is out there and AI is really helping us like make use of a lot more of that data much more quickly. I, before this, was working in wildlife conservation technology. And so we were working with scientists who, again, had all different types of data, but just like mountains of data that they had collected over their career and, you know, would either have a very like sort of tedious process for analyzing that data or they sometimes just hadn't ever really been able to like even look at all of the data that they had collected. There's just so much of it and looking at it or listening to it or watching it all manually, even with a team of grad students, was just not feasible. So I think in both of these roles that I've had, AI has really been an amplifier of the work that's already being done and really just allowing experts who are already doing really important work. to be able to like be more efficient about their work or not have to do so much of the really tedious stuff in their normal process and be able to focus on things that are, you know, the things that we really need human experts to be focusing on. So I think, yeah, it's gonna really expand what we're able to do in the environmental protection and monitoring space across the board. That's so cool. Yeah. Well, I'm thinking about just as we talked about the size of the data sets being substantial and then like people, as it sounds like, tell me if I'm wrong, that people like are potentially accidentally are sitting on gold mines of like data and like very classically, like a data lake or data moat, just like, even with a team of grad students, such a vast data set, almost unfathomable to even deep dive. Is that what you're saying? Yeah, yeah. So like, for instance, one of those projects, we worked with this scientist who, her research, she had, she and her collaborators had proven that dolphins, bottlenose dolphins make unique whistles, each individual makes a different whistle from all the others. And so she had, I think it was 35 years worth of recordings of dolphins that she had the gold mine she was sitting on, but it was just, you know, not like she did not have enough time or money to pay people to listen to 35 years of recordings. And so, yeah, AI in that instance was able to build her something that could listen to all these recordings and pull out the whistles and then try to identify the individuals that were making each whistle. And building that AI was totally like, like her research was totally foundational to ever being able to like even start to build that AI. But then building it was able to like, give her a tool that could really save her a lot of time essentially in her in her research going forward, and enable other people to utilize the research that she had done. She knew sort of how to label these different whistles as the different individuals. But AI could then help. Maybe we expand it to another population of dolphins or, um, you know, just enable others to take advantage of the really, really fascinating work that she had done. Yeah, well, and I mean, dolphin whistles, I love that use case. But also it's interesting, it's like, again, it's like hand labeling, a lot of it seems almost like, you know, these huge data sets, and going from kind of a manual, regardless of if it's actually pen and paper, to like a digitized, and then AI takes it to a whole different place of these patterns and labels, as you mentioned, like how we think about parameters, like it's this evolution that's happening so fast in front of us. or otherwise to make use of it. That's so, hmm, very cool. How long were you working on the wildlife projects? Yeah, I was there for about four years, four or five years. Yeah. Nice, that's so cool. Well, let's see. And as you, I mean, you've done this wildlife stuff, you're doing this geospatial stuff now, how are you seeing the field evolve in front of you as you're, you know, the last six years? Yeah, I think the field, well, it's something that's been sort of a, I guess a challenge that I've encountered a few different times has been like the interdisciplinary nature of this work, because we're working at the intersection of multiple fields, you know, like a lot of people in AI are. listening to those experts on the environmental side and learning from them is hugely important. And then also having those experts be open to using these new tools and incorporating them into their workflow has been challenging as well in some ways. When we go into these projects, I don't think we're ever trying to force new tools onto people. Trying to help people see that these tools really could be beneficial to them is, you know, something that takes time and like communication and trust. So I think as time has gone on, though, as more, you know, wildlife conservation experts have used started using these tools, like they've become more and more mainstream and normal and, you know, they're introduced to. their peers and things. So I think that transition is getting easier of like helping people see that these tools can help their work progress. Yeah, so I guess that's kind of an evolution that I'm seeing in the field. Definitely, and I think we're seeing that in a lot of different fields in general. It's adoption and anxiety and then trough of disillusionment and then realizing, wait a minute, there's this other tool. And I feel like it's almost functionally impossible to keep on top of it. Honestly, I was talking to a friend recently and she is trailblazing in the facial recognition space. Joy, you're on. Yeah, acquaintance, acquaintance. yeah. She was saying it's hard to keep up. And I was like, okay, if Joy has trouble keeping up, good luck for the rest of us. You know what I mean? Like, the thing is. like I can totally relate to their anxieties and yeah, you know, I think it's very, very understandable and probably, you know, we should be a little bit hesitant to just bring new things into these methods that have been working, you know, well enough. So, yeah, I think, I think it's all part of, it's very normal and expected part of the process. Definitely, definitely. I think, you know, adoption without guardrails. I mean, we saw the Samsung data breach, which, you know, scared, anyway, I think there's also, my sense, and that's my own perspective is like, this is the slowest it's ever gonna go. And if you don't jump on to the bandwagon, I'm worried you're gonna be left behind. Or like, I really, maybe, I feel like that even could be fear-mongering. I really don't intend it. I mean to be like, this is a perfect time. like jump in and keep that curiosity and that learning mindset to continue to grow and evolve. Okay, well, you are doing some amazing work. Now we jump to the career section and say, how did you get here? Yeah, how did I get here? So I have a bachelor's and a master's in computer science, but when I started college, I did not know anything about computer science or even like what that meant or what you could do with a computer science degree. I really at that point thought you had to be like really into video games and like, building your own computers and things like. to be able to work in technology. And so when I got to college, I was introduced to like this whole new world, had a lot of imposter syndrome, still definitely sometimes struggle with that. But I did eventually get my degrees and then started out in that wildlife conservation role right out of school, which was really fortunate. Like I definitely didn't plan to go into. any kind of environmental job. But, you know, I always loved being outside and like, you know, living in the Pacific Northwest, like it's the nature is like such a huge part of just our lives, I feel like for most people, because it is such a great place to just go outside and explore. And so I kind of accidentally found myself in this world. But once I started to learn about all the ways that technology is already being used in conservation and then all the new ways that we could bring it into conservation to like improve that field and improve the ways that we protect the planet. I was like all in. I really, really love working in this area. And so yeah, I did that conservation stuff for a while and then have moved into this satellite data world since then. But it's been really, really fun. and exciting. Yeah, that's awesome. Well, and you know, how many women have gotten computer science degrees? You know, this number, unfortunately, has dipped significantly. Did you have mental hurdles or other? I agree with you. It used to be like, don't you have to know video games and wear black hoodies? And you know, like, there's a perception of this. My friends who are women who got... got computer science degrees, had to be like, there are two of us in the room. Like, I just have to be okay with that. Like, how did you experience getting these degrees? Yeah, so the reason that I ended up taking my first coding class was that I was in this scholarship program that was created to encourage more women to go into math and computer science. And I have always loved math, so that's why I went into that program and assumed like, okay, I'll be a math major. But because of that program, like they had these mandatory coding classes that we had to take. And I never would have explored the field without that program. So shout out to them. I went to Western Washington University. I had like an amazing experience there. Loved the computer science department there. But yeah, so I had to take these coding classes and the first one that we had, that we took as part of that scholarship cohort, there were like only women in that class, plus a few men who, But within that scholarship program, all the women took this first class together. And we, a big emphasis of that class was to look at the ways that computer science could be used to help people and to help solve problems. And it was really framed as just this tool that could be applied to all these different areas. And that I think was like an amazing way to be introduced to. computer science, because that was like my entire goal, like my whole life was how can I get a job someday where I'll be helping people or like doing something that feels meaningful, which I think like a lot of us have that goal, but that I think, you know, there's been studies that women especially feel really motivated to go in that direction. So framing computer science in that way to this group of women was a really good idea, I think. And from there, because we had this scholarship cohort, we kind of went through our classes together. And that was also like a huge game changer. I don't know that I would have made it through into like actually deciding to major in computer science without that group of women. So yeah, the community that I was able to be a part of was a huge, like that was the differentiator for me, I would say. That's incredible. I'm guessing there was a few humans who helped curate and make that program possible. And it's so amazing to hear that shaped your experience. Or, you know, you can be full of purpose and not, there's just so many different ways to help society. Yeah, yeah. And it was two men who started that program, Dr. Perry Fasano and Dr. David Hartenstein. So they were in the math and computer science departments and they were like, you know, we see this issue, there's not enough women in our departments and let's do something to solve that. So really amazing. Yeah, yeah. Yes. Well, and you also got a master's degree, which not all computer science folks I know do that. Why did you also go for that, the further degree? Yeah, yeah. So partially it was because Western has sort of an accelerated master's program, so I could start taking classes for that and then just do an extra year of school and finish my master's. So that was really appealing. I was like, well, just one more year, why not do it? And I also wasn't really ready to like, decide what I wanted to do yet and go out and get a job. So I was like, okay, one more year would be good. I was also doing really interesting machine learning research with a professor at Western as an undergrad, and so I would get to continue that and keep working with him in my masters. And so that was part of the decision as well. And now I'm really glad I did it because I learned a lot of really applicable research. I got a lot of applicable research experience that has... applied really well to the jobs that I've had since then. So it was really a good choice for me and really helpful in my career. Yeah, that's awesome. Well, and when you think about, and this is, you know, it used to be a spicy question, I think, and now that Jensen of Nvidia has said, you know, do we need coding in the future? Like, is this a helpful skillset for everyone to know? And I don't know if you saw the demo today with Devin. It just like spits out code and does all your debugging and stuff and builds products. I signed up to be a beta user. Like when you think about, computer science, right? Like what we need to know, like the hands-on piece of it, machine learning. I mean, you gave some obvious examples of why these efficiencies are extremely powerful and helpful. I think there's, you know, where the future is heading. Do you recommend to people around you to pursue computer science degrees? Like what's the utility of, or where do you see the field of development going? Maybe it would be a broader question. Yeah, that's a great question. I do talk to students as much as I can about how they should learn to code. So I'm definitely a big advocate for learning to code. I'm totally aware that majoring in computer science is not going to be for everyone, but I think even with tools coming out that can write your code for you, I think it's still really beneficial to... learn to code and understand what is happening and how that code is working. I think no matter what career someone wants to pursue, learning to code is gonna be a great thing to know. And just give you kind of another tool that you can have in your toolbox to be able to pull out when you need it. So yeah, I still am definitely a big advocate for learning to code. But I mean. Yeah, increasing efficiency has been a huge part of everything that I've worked on. So I don't think it's a bad thing either that there are more tools that can help. You write your code faster. Yeah. no, in my PhD program, I took a course in R, because everyone was just like, a little bit more R is going to save you so much time. And now, I don't know if you've seen like R tutor, where you like, hey, I'd like you to make a scatterplot, and it pops out the R code, which is actually better than ChatGPT's code at this point. But like, that would have saved me so many hours. Or just like, honestly, I have jealousy for the people now who don't have to slog through. You still have to potentially debug or there's things that don't work the way you want them. But just the, I think the accessibility of like what that draft starts as, is it just a different place, you know, here in March 2024, right? Like, where's it gonna be next year? I agree about that understanding. I honestly, my CTO, we think about like kind of the bifurcation. of like a CTO type role of someone very senior and like the individual contributor, like trying to get in, I've heard like the first rung in the ladder is one of the hardest career-wise to be able to have one's senior engineering trajectory. When you think about, so if people hear this podcast and they're like, wow. Gracie's job is it? Like, how do I get that job? Or like, how do I pursue those type of projects? Like, what kind of recommendations might you give them? Yeah, I would say one of the first things to do, and like most immediate things that anyone could do is to go find some data. There's like all types of wildlife related data, like environmental data sets out there that you can just play with and like train a model on, you know, and learn about and figure out how was it collected, like what pre-processing happened to it and you know, what can you do with that data? So I think, yeah, finding datasets that you think are interesting and just playing with them is a good first step. And then there are also platforms out there where you can get involved in an active project that's in the conservation space. So there are sites, there's one called wildlife.ai, fruitpunch.ai, and then opensustain.tech. All three of those have, yeah, just like volunteer projects where you can... volunteer your time to contribute to a real project that they've put together. So I would check those out as well. And then beyond that, if you're ready to like get into this field, my friend Carly and I, we both on our own had kind of been realizing that there's like a lot of people who do wanna get into this field. And then also within the field, there's a lot of people doing very similar work who maybe don't know about each other. And so we created a resource called the Conservation Tech Directory. And in there, we've had the community help us create this list of just all the organizations and resources and universities and companies like doing work in this field. So you can go look at that list and see, you know, what companies are out there that maybe. want to apply to an internship or for a role there, I think that's a good place to start as well, to see the like vast world of different organizations working on cool stuff. Yeah, that's awesome. Well, props to helping out with a directory and connecting the dots. That's amazing. When you think about, you mentioned your degrees, are there specific programs? You mentioned internships? Or like, what are kind of the stepping stones if someone's starting from ground zero that you think would be helpful that you might want to recommend to folks? Yeah, so from ground zero, yeah, I think learning about kind of how AI is being applied in this field and like what are the current blockers that people are facing, like can start to give you ideas about, you know, if you're in school and like figuring out what path you wanna take, like that could help sort of guide that potentially. So there are a few communities that... I would also suggest that people join. So there's like this AI for Conservation Slack group, which has a bunch of conservation people and a bunch of AI people, and they're all together helping each other, like posting about things that they're struggling with, or like, does this tool exist or do I need to build it? So I love that group for just like learning about real things people are working on right now. And then Wild Labs, as well as a community that you can join with. again, people from like both all the whole spectrum of like wildlife conservation technology. And there are like internships and things posted there jobs are posted there. And then again, just like forums for people to ask questions and offer solutions. So yeah, I would suggest checking out those two things. And yeah, the best way to get into this field is to just like, start, start trying to connect with people and talk to people and keep your eye out for different opportunities that you're interested in. That's awesome. Yeah, I think that's excellent advice. Well, any other, you're just giving some great nuggets of advice, but any other advice, takeaways you might wanna share? I mean, I would say to just one of the most important things for me in this field has been to be really open to listening and learning from people who have different backgrounds than I do or people who are ahead of me in their careers. I think working with any type of technology means that you're going to have to keep learning and... as things are changing, and that can be really intimidating. So I think just trying to be as open as you can and, you know, giving yourself, being patient with yourself as you learn all these new things and put all the pieces together, I think is important. But this field specifically, like the intersection of AI and conservation and environmental science, like really needs a lot of... smart and motivated people. So if you are interested in joining this field, I think it's a really fun place to be in and we could use your ideas. So come work with us. I love that. I love that. And I think especially, I don't know many people in your field. And I'm here in Seattle. And it's just like a lot of B2B SaaS, like tons of software people. And I think I am, I am working more and more on like climate tech. And my mentor was like, there's all this climate stuff. And there's the AI folks. And why aren't we talking more? Like, he sees this is for whatever reason, relatively divorced. Or like you'd think these are amazing use cases and high impact. So I think this is another one where it's like this part of the field, and even as you're talking about the advances going on, but like it sounds like there are some more advances or like it sounds like a huge amount of work to be done. Yeah, yeah, and even just bringing, you know, advancements that are already available within like AI as a field, bringing those into this space and like applying them to these challenges that exist for conservationists, like is still something that is kind of a work in progress, I would say. So yeah, there's new innovations. And then there's also, let's just introduce the innovations that already exist. what, like 35 years of dolphin data? Are there, do you know of any other data moats we should talk about and like, volunteer for this data set, any obvious? gosh, I should have looked at specific ones. Oh, I will also plug Zooniverse is a really cool, like crowd sourcing platform that has a lot of wildlife data sets. that are real, you know, real scientists put them there, hoping that people will contribute to them. So if you want to label some pictures of penguins and like, you know, whale songs, like that's a place to, that you can make an impact. Yeah. fabulous. Well, thank you so much, Gracie, for your time. It was wonderful speaking with you. Yeah, thank you for having me Joan. It was really fun. Oh gosh, was that fun. Did you enjoy that episode as much as I did? Well, now be sure to check out our show notes for this episode that has tons of links and resources and our guest bio, etc. Go check it out. If you're ready to dive in to personalize your AI journey, download the free Your AI Roadmap workbook at yourairoadmap .com / workbook. Well, maybe you work at a company and you're like, hey, we want to grow in data and AI and I'd love to work with you. Please schedule an intro and sync with me at Clarity AI at hireclarity .ai. We'd love to talk to you about it. My team builds custom AI solutions, digital twins, optimizations, data, fun stuff for small and medium sized businesses. Our price points start at five, six, seven, eight figures, depends on your needs, depending on your time scales, et cetera. If you liked the podcast, please support us. Can you please rate, review, subscribe, send it to your friend, DM your boss, follow wherever you get your podcasts. I certainly learned something new and I hope you did too. Next episode drops soon. Can't wait to hear another amazing expert building in AI. Talk to you soon! Hey folks, jumping in here to say that we are so excited about how well season one is going. My DMs are full of people who actually want to be featured on the podcast, which is so cool. I love people taking the initiative to be like, wow, I heard that episode, can I be next? We would love to have more ideas of guests. I've been tapping my network and when I see people who I'm like, ooh, I'd love to hear his, her, their story. We would love to expand the network and the stories that we can share about cool AI projects, about career trajectories, about entrepreneurship. We are here for that. If you wanna self -nominate or nominate someone else, please go to yourairoadmap .com / podcastguest. We have so many. cool ideas for guests in the future. think my team, I think I may have written down like 300 guests that I'd love to have on. My team was like, wow, okay, how many seasons are we doing, Joan? I like, well, let's find out. there's so many cool stories and projects around the world. I swear every episode every guest I talk to, my knowledge of AI and the field expands and digital transformation is happening in so many different sectors in similar and different ways. Anyway, that form again for nominations, self -nominations, friend, colleague, mentor, aspirational nominations, yourairoadmap.com/podcastguest. Okay, thank you, let's go!