
EPISODE 50
Entrepreneur Experience: Carlos Gaitan, Co-Founder of Benchmark Lab
Technology has huge and far-reaching impacts on just about every facet of our world. Perhaps one of the least understood impacts it can have is in agriculture, crop production, and the associated economies that depend on agriculture. (Like, ALL of us!) In this episode, co-founder of Benchmark Labs Dr Carlos Gaitan spends time with us to talk about building an AgTech company with meaningful impacts to farmers and consumers, and demystifies AI and Machine Learning in a way that will make you feel smarter. Listen now!
We rarely have a full understanding of the role technology plays in our everyday lives. Today’s guest helps us go deep into the world of agriculture tech (AgTech) and explains just how impactful the smallest things in the physical world can be to global economies. Dr Carlos Gaitan, co-founder of Benchmark Labs, joins us today to chat about the importance of weather forecasting for prairie chickens, the real difference between AI and ML, and the importance of passion in entrepreneurship.
Benchmark Labs saw a huge potential to improve microclimates for specialty farmers and growers. Think wine, avocados, anything that’s a “small” crop size and somewhat special in how it’s grown. As Dr Carlos explains, the large crop growers, like corn and wheat, have huge fields, and thus a lot of technology companies are building solutions to help them. But the smaller, mor specialty growers, have traditionally been left with very few options for improving their crop yields and increasing their efficiency. Dr Carlos helps us understand how ironic this is by showing us how important microclimates really are when you have a smaller crop.
And did you know that the weather at the ground level can vary dramatically from the weather 40cm off the ground? Which is usually a dramatic difference from 6 feet off the ground? These types of differences can make or break a crop, and a farmer or family business, and it’s exactly the kind of problem Dr Carlos is familiar with and has a passion for solving. He also dives into the technology that’s used to solve these problems, including the difference between AI and Machine Learning and what ML can really do. (Hint: We’re not worried about real-life Terminators… Yet.)
Check out Dr Carlos’ recommended resources:
“Rotten” – Netflix documentary about avocado water needs
“Machine Learning Methods in the Environmental Sciences” – William W. Hsieh
“Venture Deals: Be Smarter Than Your Lawyer and Venture Capitalist” – Brad Feld, Jason Mendelson
Find and follow Dr Carlos on Benchmark Labs or send him an email at carlos@benchmarklabs.com.
Be sure to like, share, and subscribe to Precursa: The Startup Journey on your favorite podcasting platform and tune in for the next episode!
Email us with any questions or comments (startup@precursa.com). Check out our website (https://www.precursa.com) for more information on getting your startup rolling.
(00:04):
Straight to you from Denver, Colorado, this is Precursa: The Startup Journey. We share the ins and outs of building a tech startup from inception, to launch, to revenue and beyond. If you’ve ever wondered what building a startup from scratch really looks like, you’re in the right place. With full transparency and honesty, we reveal it all about Precursa on our ride from idea to exit: the wins, the lessons learned, and the unexpected twists and turns.
(00:37):
Hey everybody. And welcome back. This is Precursa the startup journey. And today I am joined by Dr. Carlos Gaton, who is the co-founder and CEO of benchmark labs. His background is environmental sciences and machine learning, and he’s an elected member of the American meteorological society’s artificial intelligence committee, which is so freaking cool. And we’re gonna talk about that. And it took me forever to figure out how to say that without like flubbing the words. So there <laugh>, he’s worked at early stage startups in the environmental science space, and he’s here today to share his entrepreneurial journeys with us. So welcome to the show, Carlos,
(01:21):
Thank you for hosting me, Cynthia. Eh, very excited to be here.
(01:24):
Yeah, I’m excited to have you, so start by telling us a little bit about yourself and how did you become an entrepreneur?
(01:31):
Yeah, thank you. Yeah, that’s uh, super interesting question. I guess, that I was always entrepreneurial since, uh, growing up all of the situations that I had lived in the past from my academic career to my personal experiences, uh, put me in the situation that I’m here now as a CEO of benchmark labs. So yeah, as you mentioned, have a background in atmospheric sciences, my machine learning, but also I think that’s important, uh, to mention I, when I was growing up in my native Columbia, we experienced a very severe, uh, event effect that, uh, created, uh, or affected the whole country. We were highly dependent on hydropower. We have lots of rivers, so the country traditionally had relied on dance and hydropower, but the, that, uh, linear event was so severe than we even have to move the clock one hour, eh, Columbus located in the equator. So imagine like having a winter savings time or something. So that was, uh, that was shocking, uh, for everybody including myself, I was studying. And, uh,
(02:42):
How old were you when that happened?
(02:45):
I remember I was doing, I was doing homeworks. Uh, I remember like I had to sometimes do candlelight <laugh>. Wow. And, uh, so that’s, uh, yeah, that was very, very interesting. And since then, it’s like, oh, I have to study all these effects of water, climate weather, how it affects economies. So, uh, then I did a hydrology hydro informatics. Uh, then I went to, uh, the university of British Colombia to do my PhD, uh, with professor William, she that, uh, one of the pioneers in machine learning applications to environmental sciences. Yeah. Um, because I said, you know, I study hydrology, but I want to understand what happens at the cloud level <laugh>. Yeah. So I guess that all of those, uh, situations, my curiosity, and, uh, keep understanding the water cycle and the impacts of weather all over the world. Then after that, I moved to the United States.
(03:42):
I went to Princeton to geophysical Floyd dynamics laboratory, uh, worked there in close collaboration with scientists, from the department of commerce department of interior stakeholders. And, uh, they, since that moment, it became very clear that, um, I had like also experienced more joy, uh, working with users, solving problems. Yeah. Um, for different stakeholders than writing papers. Uh, so then I moved to the private sector, working power companies, climate consulting companies, and then we created benchmark labs to solve all those needs that we have been hearing for the last 10 years, uh, of real users saying, Hey, Carlos, what can you do to help us with weather forecasting for our location? I only need better weather for my farm. I don’t need better weather for my county, my
(04:35):
Right
(04:36):
State. It just give me better forecast for my farm. So, um, that’s uh, how we ended up here. <laugh>.
(04:43):
Wow. That’s amazing. So you literally had this, this, you know, huge life changing event, like, you know, world economy changing event happened when you were a kid and you, you kind of had this moment where you were like, I don’t like this feeling of powerlessness or I don’t like not understanding all this dynamic. I’m gonna go like, learn about that. And now you’re helping like thousands of people. Right? Like tell us a little bit more about what benchmark labs does and how you guys are solving problems and who you work with.
(05:19):
Yeah. Yeah. Thank you. Uh, so, so yeah, that affected us in so many ways, the, the drought not only affected the energy sector. Yeah. And that had like all implications that we have to go into energy setting, but we also having to save water, eh, all the crops, eh, had to be more efficient or like some of them were like actually not receiving enough water to, to produce at maximum yield. Yeah. So, so yeah, it was, uh, a very interesting, uh, situation.
(05:54):
<laugh> <laugh>. Wow. That’s really cool. That’s really, really cool. So tell us what you know, you’ve been, how long have you been an entrepreneur?
(06:06):
Good question. In Colombia. I started like, uh, probably growing up and after being 18, probably I did a lot of things. Probably. I even sold sandwiches, uh, with some of my friends. Oh,
(06:17):
Wow. No
(06:17):
Kidding. And, uh, time to get like a then, uh, started, uh, doing some, uh, consulting when I was doing an engineering, I did a, a blueprints, uh, help with, uh, automated software, uh, to I always to provide some services. So I had like some idea of customers and making value added solutions. And then, um, yeah, I guess after the, I went to Canada for my PhD, I put that on hold then after moving to the United States. And after my experience, uh, at Noah, when I moved to the private sector, then, uh, in 2019, yeah. We decided to create a benchmark labs. Okay. And what we do in benchmark labs to answer your, your previous question is like we do, uh, environmental forecasts. Okay. Tailored to each specific, uh, Pharmac each specific field. Wow. Uh, so what we do is we integrate IOT sensors that are like internet of things, basically anything that is connected hardware sensor that transmits environmental data, uh, through the cloud. Yeah. Uh, we improve the forecast from the national weather agencies from different private providers. Yep. We correct those one to be more actionable and, um, kind of reduce the errors. Yeah. Those forecasts based on information of the, in C two hardwares that are installed, that, uh, style that, eh, fixed assets like agricultural farmers.
(07:51):
Wow. That’s super cool. So you can tell a farmer who’s, you know, got 200 acres of cornfield. It all starts with your idea, scratch that your great idea. So you do your homework in, because you’re a doer. You make a plan, you raise the capital, you find a good developer and boom, your app is born, but even the best plans for these great ideas rarely turn out. So linear testing, bugs, user feedback, and unforeseen setbacks can make an expensive mess of things. Did you know that on average, you’ll spend more than $600,000 over 36 months to realize zero revenue. In fact, in 20 18, 40 6% of startups failed because they lacked the experience and skillset to successfully navigate this challenging entrepreneurial journey, even worse, 42% of these great ideas failed simply because there was no market for the product in the first place. The good news, there’s a better way.
(08:53):
Precursa provides qualified, specific, experienced feedback from those who have taken this journey before. That’s the kind of informed research Google can’t provide. Precursa provides a time tested sequential roadmap, meaning you’ll always know the answer to the ever present question. Now, what, and Precursa has successfully navigated the stressful turbulent, but necessary steps to start up success. So when you’re ready to take the leap, your roadmap to successful launch is more direct with far fewer pitfalls. We believe entrepreneurs like you change the world and we provide you with the best tools to get there. Like very, very accurately what to expect for what five days, seven days, like how far out are you forecasting? Yes.
(09:42):
<laugh> yes. So, um, we focus on, uh, climate sciences and the NOIC sciences is called the weather window. So it’s like what’s happening now. Okay. Up to 15 days in the future. Oh, wow. Wow. And, uh, we improved the forecast from national weather service up to 60, 66% depends on the geography and the valuable that you are improving. But, uh, farmers have been receiving improvements in product humidity, temperature, wind speeds, direction, even so interesting. S like ABA, transation.
(10:19):
Wow.
(10:19):
That is how much water the planter uses.
(10:22):
Wow. That’s cool. And how many sensors do you have to put in order to get this data? I mean, is it like you have, you know, a couple and some weather veins or, you know, how many, how many does the average average installation have,
(10:39):
Uh, to deploy our technology? The farmer only needs one sensor,
(10:43):
One sensor
(10:44):
Information for data specific location, some farmers, um, if they are not into row cross, for example, in specialty cross, they, uh, rely more on effects of micro climate. Their, there are as the French call them, they value of what makes them unique in terms of soil characteristics, uh, land cover, uh, interesting in micro climates that could bring, uh, sea or, uh, other node of flavor to the crops. So in those cases, uh, we had, uh, even installations where they install one sensor on top of the other. Uh, so it depends crop specific solution. Okay. Uh, it depends on the level of microclimate data they are experiencing. Okay. And how much accuracy they want to, uh, to get for their specific locations. So for, uh, road crops probably will be, as you said, like one sensor for 10 acres, maybe it’s representative. Okay. But if you are, uh, thinking about, uh, wine grapes.
(11:49):
Yeah. Vineyard, especially, especially if you’re in like a more hilly area where some of your grapes are growing up and some of your grapes are in the valley,
(11:57):
That could be pretty different.
(11:59):
Right. Some of our users have like, uh, these IOT sensors in the valley, as you mentioned, and other ones have in the valley and on the hill side of the mountains, sometimes multiple, uh, sensors, depending on which side, because they could receive shade at different times of the day and, and so on.
(12:16):
So, wow. That’s so cool. So you’re building machine learning models and artificial intelligence models to help take all this data and do higher accuracy predictions. Am I getting that? Is it, is that an accurate assessment?
(12:32):
That’s correct.
(12:32):
Okay. Okay. So there’s a lot of myths, I think. And, and a lot of like believed things that aren’t necessarily true by, by the general population and even a lot of people in tech about what is AI and machine learning and what can it really do. So can you tell us more about like, what really is machine learning? What can it do? What can’t it do, like break, break some of that down and, and let’s dispel some myths about that.
(13:04):
Yeah.
(13:05):
Are our machines,
(13:07):
Are machines gonna replace humans? Like, that’s the big question, right?
(13:11):
<laugh> yeah. We, we also Terminator and know how we did, so
(13:15):
<laugh>
(13:16):
It’s uh,
(13:16):
<laugh>
(13:18):
Exactly.
(13:21):
So, yeah, that’s, uh, let’s start like, um, machine learning and AI sometimes it’s they are being used interchangeably, but machine learning is a subset of techniques of AI. Yep. That is like the broader field of artificial intelligence. In this case, what we are implementing it’s, uh, machine learning techniques where the, um, the, uh, algorithms learn as more data comes, uh, we constrain them by physics. Uh, so they don’t go and produce something that is physically
(13:53):
Not possible.
(13:54):
Environmental science.
(13:55):
It’s gonna rain diamonds in Iowa today. <laugh>
(13:59):
Exactly. Or like,
(14:00):
You know, say like relative humidity is going to be 300%. It’s like, that is not possible, but sometimes the computer doesn’t know. So, uh, so probably eventually, maybe AI in the future, I’m saying like, they would be intelligent enough to, to do that. But machine learning just, you have to constraint. It’s more like they are based in the statistics to, to remove all the meats around them. Uh, lots of linear algebra, matrixes, operations, uh, how to make those, uh, equations, uh, they result them faster. They, they find ways to, uh, find patterns yeah. In data. Uh, you can, uh, there are different techniques. Some of them, uh, are very good in classification. Okay. So it’s a typical cat versus dog yeah. That you see in the internet, or like, just try to see if you are human, like identify which ones are the cross, the, you know, traffic lights or the bicycles in the <laugh>. Oh yeah, yeah. You know, these, uh, so, so yeah, actually you are actually helping those algorithms to be trained because now the humans tell them, oh, actually this is the,
(15:10):
Oh, interesting. Right.
(15:12):
So, so some of them just learn by pattern. It’s like, they need lots of data. Yeah. And the more data they have, they say, okay, this is definitely a dog. This is definitely a cat. Yeah. And they, they can generalize when they see a totally different let’s say hos.
(15:25):
Yeah.
(15:26):
They could be
(15:28):
They’re like, is that a box? No,
(15:30):
<laugh> yeah, exactly. <laugh> so sometimes it’s just trying to train them at that level of, of subtlety, but they require a lot of data, ideally at the end, in, in much in artificial intelligence, then, uh, the system will learn by itself too. Mm. Um, and, uh, we will be able to, uh, treat a human to, so a human will not know that is, uh, you know, that you’re talking with a machine. Yeah. Uh, still you still, uh, there are many instances that you will know that. Yeah. Oh, this is a robot. I just want to talk to a person. Yeah. And, um, but yeah, so basically to remove the statistics and people had been doing statistical corrections for, in engineering for decades
(16:20):
Millennial, right?
(16:21):
The same thing. Yeah. Yeah. Millennial, we are used to linear regressions and the limit, these, uh, machine learning models, the most simple ones, you can, they will solve a linear regression, but then they can also learn more like non-linear regressions. Yeah. A little bit. And you can kind of aid them yeah. And tell them, oh no, don’t extrapolate too much or extrapolate beyond these bounds. Or I expect that the answer will be linear. Yeah.
(16:52):
Yeah. So it’s
(16:53):
About you also, you need some domain expertise to, to make them, uh, more actionable and, uh, realistic. And I think that some of the dangers that probably helps about the myths and the misuses of AI is when people, uh, train them without, uh, understanding the, the physics or the, the causality, what, what is beyond the problem.
(17:17):
That’s interesting. So, so it sounds like with machine learning, it, it actually requires a smart human to be able to look at the models and say, whoa, whoa, because we’ve heard about machine learning models that go rogue and they start, like, they get some kind of data in that is a little bit different or may, you know, maybe they make a different assumption and nothing corrects them about it. And so they go, oh, well, now all of this is this. And you’re like, whoa, whoa, whoa, wait a second. Hold on. Right. And so machine learning is really more about teaching a computer, a model to he, to improve identification for workflows or identification of data patterns. But it still requires that human intervention to say is the, is this, I’m not gonna say logically, but is this really right? You know, and sometimes you can only get that from knowing what you’re looking at. And, and, and like you said, having that domain knowledge. And so then the, the extension of that would be artificial intelligence as a machine that can re replicate a human in, looking at what a, what a machine learning model is doing and SA, and, and making those determinations and, and having artificial intelligence, keep a machine learning model from going rogue. Right. And that’s like way future sci-fi right. I mean, are we there? Does that exist?
(18:43):
Not there we not, we are we’re going then. Or people are at the research level actively looking into making more intelligent machines because they can make decisions, better inform decisions. Yeah. Learn like a experience, you know, like the toddlers it’s important for robots yeah. That they can learn as they go on the field and, or different situations. So there is a component of, um, supervised learning that is, uh,
(19:11):
You
(19:12):
Supervise
(19:12):
Them, okay. Teach
(19:13):
Them. But there is also a, another subset of models that are unsupervised and you give them data and they try to, it’s kind of super unsupervised, but the human tells them what they’re expecting to. Okay.
(19:28):
It’s like,
(19:28):
Okay. They say you have a lot of, uh, data yeah. Of, uh, it could be articles. It could be news articles. Yeah. And you can tell these unsupervised, uh, model to try to find, uh, three categories or five categories. Okay. That reflect all the articles on your podcast or your, or the New York times. Interesting. So the, the, the machine learning model now is constrained by the human and say like, oh, you want me to find categories? Yeah. And then they will put them together and tell you, I found five categories or seven categories. And these are the, and these are the ones that are closer to this category than to the other one. Yeah. So that’s unsupervised, but the human told them, Hey, I need seven categories. I need five categories
(20:10):
Constraints. Right.
(20:11):
And they are doing it by themselves. The other alternative is like, you have to train them and say, Hey, this is a cat. This is a dog, eh, or this a linear regression, or this is what is expected. This is a temperature, uh, here. And, uh, but if in reality it should be here. So there are, there are regression models, their classification models. Okay. And, um, that basically those ones will cover most of the applications. And in, in the future, ideally, um, you know, we have, uh, um, systems that learn by themself, uh, like more like a baby <laugh> or
(20:46):
Us, so terminators coming, but we’re not there yet is what I’m hearing you say.
(20:53):
I hope that smarter, not the keys. So Skynet to a robot that AI should control our nuclear systems, but <laugh> <laugh>
(21:09):
Oh, that would certainly change the rushy crane conflict. Wouldn’t it? Oh my gosh.
(21:13):
<laugh> yeah. You always, you, I think that you will need a human in the loop.
(21:17):
Yeah, no kidding. No kidding. We’ll call that one highly supervised. <laugh>
(21:21):
Highly supervised.
(21:22):
I, yeah, that’s, you’ve done lots of different entrepreneurial things and, and we could talk about AI and ML for, for days, because this is fascinating to me. And I’m so immersed in this right now. And I, I just love the way you broke all that down, but I wanna go back to the entrepreneurial side. So, so tell us a little bit about, like, what do you think is the most important lesson that you’ve learned as an entrepreneur so far in, in the various things that you’ve done in your ventures?
(21:51):
Oh, that’s, um, that’s a very good one. I, I think that, uh, the most important part is, uh, to, eh, listen to, to your users and create a product that actually, uh, is going to be needed or that it’s also a real clear, uh, problem. Sometimes you divide it in categories, like what is a must to have, what is a nice to have yeah. Right. To find those ones that are must yeah. Both probably serve a need, but, uh, if a user must solve that problem yeah. To be more efficient. Yep. You are not good track to, to define a product, to, to build something that it has a real value that they can spend on the, yes. I can spend this today. Yeah. Instead of, uh, having to create something that, uh, it could be a nice to have.
(22:42):
Yeah. And you kind of have to talk people into buying it. Right.
(22:45):
And then, yeah, exactly. So, yeah, I think that that’s, that’s probably the most important part. Try to understand your users and understand what is, uh, what is a real need versus, uh, what is a C <laugh> needs from the entrepreneur. So don’t put your own judgment, go there and talk to them without preconceptions. Uh, because sometimes the answers are as simple as what I really need yeah. Is to get all of these 20 different sources of data yeah. That are ed into
(23:19):
20, in
(23:20):
Single place pages. If you can just put them together. So I can say print screen, and then suddenly I, you make my compliance easier, you know?
(23:29):
Yeah. It’s yeah, yeah. Probably,
(23:30):
You know, it’s, it’s a value added, uh, solution for us, but, uh, or in that perspective, but it, it comes from listening to the user, you might think like, oh no, what they need is another webpage that is totally new or whatever. But, um, the big problem for them is that there are too many web pages. It could be <laugh> as, as an example,
(23:49):
Sort of like how we used to have cable. And then we were like, I just want the channels I want, so now you have all these streaming services, but now you have to buy all of them. And so now I’m hearing people say, I wish I could just buy like one thing and get all the stuff that I like. I’m like, wait a second. Are we back to cable? We just like evolved cable. What just happened? <laugh>
(24:09):
It’s, it’s an interesting point. I was wondering like, if you add all these, uh, streaming services that you’re paying. Yep. It seems that they’re cheap. Yeah. But if you add them all the subscriptions probably cheaper to get cable
(24:22):
That’s right.
(24:23):
To get on the customers.
(24:24):
That’s right.
(24:26):
<laugh>
(24:27):
So, okay. So I heard you say something that I’m like, oh my gosh, like a data science guy, like a, like a, an engineer said, talk to people like really, I gotta go talk to people. Is that really what I have to do? And how do you do that?
(24:41):
<laugh> yeah. It’s it is, it’s important. It is very important. So, uh, <laugh> I guess when, when I was a <laugh> it’s so it is so important. It’s not for everybody, not many people like to talk to people <laugh>
(24:56):
Yeah. Especially if you’re an engineer, right. You’re like just gimme data, but you’ve figured that out pretty quickly. Right. You were like, oh, academics, writing papers. Not really my jam. I really do like people. Right?
(25:06):
Yeah. I really like people. I, I enjoy writing my papers and, uh, wrote like chapters and really like, very interesting contributions, but talking with users and having a user to tell me, Hey, Carlos, I just need information for, for Texas. Yeah. For this area of Texas. Yeah. To help the lesser Prairie chicken, they nest at 40 centimeters above ground. And you give me forecast 40 centimeters above ground. And the answer and I was working, the government was like theoretically possible, but we cannot do it because the systems are already designed to give you forecast at zero centimeters or at ground level or at two meters. And, uh, oh, wow. And every time that we run the models, we have to do it for all the entire world. So I cannot give you the specifics. I will run south in Texas, in this county for your, for saving the lesser Prairie chicken at 40 centimeters. I would love to do it, but I can’t with these resources. So I basically, after hitting all of those interactions, talking agriculture a yeah. I have a farm near the mountains. And, uh, they’re like a unique micro climate. National water service is substantially bias. Sometimes they are eight degrees faring height in errors versus what we observe at the farm.
(26:25):
And eight degrees can be a lot depending on what you’re trying to do. Right. A lot. Yeah. It’s a lot.
(26:29):
Yeah. Especially when you have to deal with frosts. Oh. You know, like it might not be a problem when you are around like, uh, you know, 60 to 68, maybe that’s alright. But if
(26:40):
It’s different between 30 and 30. Yeah.
(26:43):
Yeah, exactly. So it’s like, okay, are these guys going to suffer a frost event? And I could lose half of my crop or not 80 degrees. It’s a lot, it’s a lot the same thing. On the other side, when you talk at the, at the opposite side of the spectrum, eh, ocean regulations are now preventing labor, uh, in the fields. Wow. Around 95 faring high, 100 faring high. They are because of course it is very important for the, you don’t want people just fainting with here.
(27:13):
Yeah. Heat, exposure, and exhaustion and all that heat
(27:16):
Exposure exhaustion. So, uh, eight degrees will make a difference is the difference between being at 100 or 92. Yeah. And in 92 you can perform labor that’s with
(27:25):
Vegetations
(27:25):
That’s, but 100 it’s like, you should be inside.
(27:29):
Wow. You never, I mean, okay. And so you said something else, which is, you said, so the Prairie chicken and, and 42 centimeters does zero, you know, ground level versus 40 centimeters. Does the weather change enough between those two? That, that can make a difference?
(27:45):
It does. It does.
(27:47):
It really,
(27:48):
It does, especially because the problem is that zero or ground level, they recorded the temperature of the surface. That is, uh, soil temperature. Oh. But at 40 you have already air, you have atmospheric mixing and yeah. It’s totally different material. So it’s like the same thing. Like, uh, when you, uh, touch two different colors and on a sunny day or, uh, heat capacity as they call it of like of solids is different than the heat capacity of air. So, uh, there are biases and there are different, uh, you know, those biases have an impact on the nesting habits of many species, leather chicken. That was an interesting example that I will never forget in my life, but it, it has a, a huge effect, also more key to outbreaks and wow. Many different, uh, applications, strawberries probably require temperature that are closer to the surface. Cause they they’re,
(28:44):
They grow lower.
(28:45):
So close. Yeah. And they grow lower. But if you have a, a trees could be tea, coffee, eh, you actually need a gradient of temperatures below and above canopy. Wow. So, uh, and the canopy of the three is not at the level that the national weather service gives you. It’s not a two meter, you know, it’s, it could be one. So, uh, you need that level of flexibility. The users are very different and their crops could be at, you know, at, almost at surface level or it could be one feet could be two feet. It could be, you know, for forestry applications, they might even require temperature at, you know, six feet, nine feet easily. So it depends. And, uh, the pests also, especially now springtime,
(29:34):
Yeah.
(29:35):
Into summer, they attack, uh, the different crops. They attack different, uh, plants in general. And you can, uh, link the attack of the past to environmental conditions really in essence, they, yeah, yeah, exactly. They, they, there is like a fantastic line of work of biologists all over the world that can link, uh, let’s say powder remind you or a different leafblower insects, uh, to environmental conditions develop humidity, temperature. So any, any discrepancy or like a big arrow, like eight degrees FHE in temperature will have also an effect on how you forecast your pests. And that might lead you to think that you are not under stress, but maybe the pest has been eating your crops for one week, which had made the difference between a higher yield or no yields.
(30:27):
<laugh> wow. See the, see, this is fascinating. This has a real life implication, not just for the farmer, the vineyard owner, whatever, but also for what crops get created, thus, what food is in the supply chain. So how did you get interested in this? I mean, it’s fascinating to me, so it’s no surprise that you’re interested in it, but like, I mean, obviously, you know, you told the story about, about how you grew up, but sort of, how did you, how did you end up here? I mean, I mean, what, what was it that you were like, oh, I gotta go solve this problem.
(30:59):
<laugh> yeah, no, it’s, uh, as I told you, probably the, all the addition of all my life events and, uh, who I am talking with users and last, uh, growing up, my mom worked 28 years at the national Federation of coffee growers in Colombia. Wow. So, uh, so yeah, for us coffee was a big thing. And we even remember when there was like a, an insect coming, uh, from Brazil. I remember that it was at the, we were monitoring even at the news level. That’s like, okay, this insect is going to come to the country. And then 30 years later it’s been in the country and it’s very hard and this is still attacking the, all the coffee plantations. But, uh, yeah, it’s all this effect of how you put together, uh, the effects of weather climate into agriculture, into energy. I have always been fascinated basically, cause I grew up like following coffee, that is a very premium specialty crops, typical characteristic, specialty crop, uh, that has a huge effect on national economies.
(32:07):
So, uh, in many ways I also want to create a company. We are striving to go there to be the go to solution for a specialty crop growers in the world. We see that most of the products out there, uh, are tailored to the row crops that we have been talking at the beginning of the program. Yeah. Think about the weeds, the corns, those are like hundreds of thousands of acres and that traditionally they have like more agricultural insurance. Uh, if you lose 100 acres, uh, course it’s catastrophic, but it might not be the end of your farming life because you have another 1000 acres. Yeah. Uh, so,
(32:51):
But if all you have is 50 acres or a hundred acres and you lose half of that, that’s bad. Yeah,
(32:57):
Exactly. So it’s, uh, um, that’s why I started like working, uh, towards solving problems for specialty crops. Of course the solutions are very similar for road crops, but uh, deeping into my heart and who we are as a company. We want to make these technologies more accessible. Yeah. For, uh, medium size farmers in the specialty crop. So for example, working here in California with avocado growers. Okay. Um, we recently are starting a pilot also in Chile, no, a Chi company, uh, that has operations in, in south America, um, that they grow avocados. They it’s fascinating how avocados use so much water really. And, um, yeah, actually there is a very interesting documentary in Netflix that you should watch it. Okay. Uh, about it, uh, avocados and how much water they use. Uh, it’s, it’s very, very interesting. And here in Southern California, I read the news, uh, probably two months ago that, uh, for policy regulations, of course we are now never ending drought,
(34:02):
Never ending
(34:03):
Water districts have to make decisions of who uses the water and increasing the, everybody has to many ways, uh, is, uh, a higher cost of water than they used. Yeah. Eh, now there are restrictions on use restrictions last year. Some areas didn’t or actually were going to be paid for not growing crops.
(34:26):
Wow.
(34:27):
So the situation is, is verys
(34:29):
Very, so if you’re a avocado grower in Southern California or anywhere in California, right now, you’re struggling to figure out how to get enough water to produce a yield. And in some cases being incentivized to not grow,
(34:43):
Yeah. Some of them are facing the decision is like, can we have these? And we keep doing what our farming has been doing wow. Uh, for decades. Um, because there are so many different factors from wow, uh, imports of avocado to how much average consumer is willing to pay for avocado. Are you, it’s fascinating at the policy level, and this should open an interesting discussion of how much is the user willing to pay for the school cost of agriculture, because there are so many subsidies yeah. That are there that are hidden and of course they’re needed for national security. Yep. But the cost that, for example, somebody will pay to regain their loan is, uh, probably $2,000 more where acre feet acre feet is a unit of volume of water. Yep. And, uh, some, um, water districts, um, let’s say I could pay 2060, uh, dollars per unit. Yeah. Uh, at a residential area, but uh, in a agricultural district they could pay only 60. So, you know, it’s wow. It’s the same amount of water
(35:52):
<laugh> but the perceived value. Wow. Isn’t that interesting. Like we, we value more having a green lawn than we do paying farmers and growers for a fair wage for what they’re actually growing and producing that we eat.
(36:11):
Yeah. I say it’s, uh, it’s a fascinating topic. Um, cause it could have an effect for the government is doing that to try to keep the prices down
(36:21):
Of course. And
(36:23):
Keep people doing farming is very important. So they know everybody moves to the cities. Uh, so, but it, it puts into question, what is the real cost of, uh, strawberry yeah. Or, or avocado. Yeah. If you remove the subsidies, so how much water you need to produce one avocado.
(36:41):
That’s right. That’s right. And, and it makes sense because disproportionately, um, the poor and the lower classes are disproportionately affected by increases in food prices and increases in commodity prices. Right. Then, you know, then the average middle class or upper class person, you know, definitely in America, but certainly other places as well. Yeah. Yeah.
(37:05):
Now with the localization and I say it’s a very important effect, uh, that everybody suffered before countries were more, uh, reliant on natural or like their own economies. Yeah. Uh, now we ation, we even see some of the unfortunate effects of course. Of the Russian and creating conflict. Yeah. Um, just, I think that, of course at the human level cultural level. Yeah. But if you look at the agricultural sector overall, we see, uh, how reliant yeah. Many economies are of, uh, Russian fertilizer. Yeah. Um, the fertilizers also were coming from China and China decided to not export as much as they were doing before. Yep. Uh, then, uh, the, all the cereals produce in Ukraine. Yeah. Sunflower oil. Um, many African countries were relying on those products. Yeah. Even, uh, talking with some farmers in Colombia, some of them were mentioning that cause of the rising cost of fertilizers. Cause there’s not enough now, uh, supply, they, those costs would have to be passed to the end user or they would have to feed about other options
(38:24):
To do
(38:25):
Something else. It’s complicated. Wow. It happens globally. Wow. And for the smaller, for the smaller farmer that’s of course, as you were mentioning, it’s a disproportionate effect on bottom line and that could all of these chain of events like, uh, more weather volatility or variability, extreme events effects on labor increasing cause of all regulations. Wow. Uh, lack of water availability. Yeah. Plus all the external, uh, aspects of, uh, uh, international conflict is affecting fertilizers. And, uh, the cereals that sometimes are needed for, uh, animal feed and actually affects everything cereal is, is, is used globally for so many applications. So, um, so yeah, they, they sort
(39:15):
Of seen, so every little thing that you can do with your sensors, with the data that you’re collecting to help farmers be more efficient, it really does make a bigger difference than maybe it might seem on the surface because there’s so many factors they can’t control that if they can get really efficient with the ones that they can, then it makes it easier for them to offset some of those other costs or at least absorb some of it or, or, you know, be more efficient in other places where they don’t have a lot of choice. Right.
(39:46):
Exactly. Exactly. Yeah. Um, I think that, uh, that’s probably the, the biggest, uh, um, moral of the story. Yeah. Uh, that there’s so many things that you cannot control. All of those will affect agricultural users lower. Uh, and although, uh, we cannot control the weather to, to an extent we can monitor better and we can understand what is going to happen in the next, uh, few days at that farm level. And that has implications for labor scheduling. Yeah. For a planning when you have to do harvesting filling. Yeah. Uh, sometimes decisions as, uh, innocent. Uh, should I open the, the sprinklers or the, the irrigation water system? Uh, many users have a fixed schedule, let’s say Monday, Wednesday and Friday. Yep. X amount of gallons per minute.
(40:38):
Yep.
(40:38):
But what happen if modern nature is going to give you that resource for free?
(40:42):
Yeah.
(40:43):
You know, it’s like, it should be more intelligent at this day and age. Yeah. Uh, same situation with fertilizers fertilizers. If you apply them before it will rain, the fertilizers will grow in runoff and they will end up in a river stream and the has more ecological problems because then the nitrogen will end up in the readers. Right. Then you will have LGI blooms and,
(41:08):
And it’s not columns. It’s not helping your, it’s not helping your crowd and it’s not helping, it’s not helping the environment.
(41:14):
Exactly. Oh man. At the end you spend it, you spend the fertilizer, but the plant didn’t use it cause you have time to do the nitrogen observing. So, uh, all of those is more like decisions, you know, if you just know that, okay, it is going to ring tomorrow, then I don’t fertilize today. Yeah. You can just wait after, uh, you have a dry window, the same thing with pesticides, uh, for those ones, those farms that they still need to do pesticides. They add windows of very clear that the manufacturers have in terms of, uh, wind speeds that they need in order for it to not blow away.
(41:50):
Yeah.
(41:51):
So many, many decisions they can be just wasting money, uh, because they apply it when it’s too windy. Yeah. Or if it’s going to rain the next day or the same thing with the, with the water availability. Yeah. So, uh, if you just understand the future a little bit better. Yeah. All of those decisions over a growing season, they adapt
(42:10):
And it’s not about the macro energy. It’s not about the macro Fu you know, future it’s about for my area right here for your area, which may not be reflected. I mean, how many times have we seen that, where the weather says, oh, over Denver, it’s gonna be, but I’m in Southwest Denver, which means I’m always gonna get more snow. I’m always gonna get more wind. Like it’s always gonna hit me first. And exactly. You know, Denver’s a big place. Right?
(42:34):
Denver, Denver is a, is a big place. And that’s, uh, exactly the, the problem that people were facing. Yeah. So traditionally these weather models were created after the second world war, uh, for different applications, you know, for say based on physics, but it was a problem of national security of like when to send the planes and all those things.
(42:55):
No kidding.
(42:55):
But they were, they were models were not designed to produce the best forecast for agriculture farms or any fix asset. So, uh, we have been using it because it’s a fantastic technology. Yeah. And, uh, we identify more than sent to use cases, uh, or business cases of where applications. Yeah. But in the case of fixed assets, things that do not move in a space like a farm. Yeah. And in very human nature, you care about your farm and then you care about your county. Yeah.
(43:24):
So,
(43:25):
Um, so as you say, like if you are in, in south Denver, you understand the fact of the Rockies, depending on which side of the mountain you could see more snow noise. No. Yeah. Same thing with the, the lakes, the great lakes in the Northeast, uh, why Canada might be covered in snow. Uh, both of them might not, uh, because even the, these great lakes have an, an effect that they changed micro climates. Wow. Uh, we have seen, uh, changes in, in weather just because of, um, at the farm level. Yeah. Some farms, for example, have a crops that are growing and, uh, another field near there basically same farm. They, they are just, they just finished harvesting. So in one you have like dark soil and the other one you have like green, uh, vegetation growing. And that by itself changes what you call the albido how much energy is reflected and so on. Oh my God. At the farm level, you, you can see 2, 3, 5 degrees. Uh, difference just is almost the same micro climate, just different land cover. That
(44:28):
Is fascinating. Oh my gosh. So company, so, so you saw this opportunity to come in, you know, with all of your background and all of your knowledge and all of your, your, and I, I just, I get your heart for the climate and for people and for, and, and you have to hold these like really seemingly different paradigms all at once when creating what you’re creating, which is this isn’t just about, it’s about micro climates, but it’s also about micro economies and the impact of that and how, you know, bringing all this kind of stuff together to give these people data that they need. And, and it’s also interesting that you brought up the thing about how the, how the weather system was created, because there’s an inherent bias in, in that weather modeling for a totally different type of activity than what the average daily user really needs weather for. Right?
(45:26):
Yeah, totally, totally.
(45:27):
Which is also one of the potential side effects of, of a, of a machine learning model. You know, we see this in medicine all the time where it’s like, oh, you know, BMI, which is supposed to be like ideal body weight or ideal body mass or whatever. It’s like, well, but that was developed in Europe, in the twenties and thirties and forties with primarily white men. So like, it doesn’t reflect most people across the planet, but we’re trying to fit inside this box. Right.
(45:53):
Totally. Totally. Exactly. There are so many unconscious biases, even when you train models and that’s another situation that we can go as deep as you want, but there they are. Now there is an emerging field of ethics in machine learning. ITing try to, uh, even address or acknowledge that there are biases. Yeah. As you said, like, for example, when disproportionate at the beginning, they were like training facial recognition algorithms. Yep. And, uh, they train them always with white men.
(46:24):
Yeah.
(46:26):
So what if you are not white, you’re not right. You’re not a male.
(46:30):
Right.
(46:31):
They, they map you into what an animal or yeah. <laugh> they say like, what are you, you know, like, it’s like, it’s not, so it’s a, it happens. And it’s just cause of the, the sampling problem of, of that, the same thing with the body weight. Like if they think that, uh, European men in the, in 1920s, <laugh>, it is of 7 billion people is like, what is happening? Yeah. You know, it’s what happened with women. Yeah. You know, and it’s totally different anatomically. So it’s, uh, you can just not have like one size fits all. Yeah. And that’s why we decide also to do these solutions that we understand that we cannot give them a one size fits all where solution we are tailoring it to each specific farm. I love that specific field.
(47:21):
I love that. All right. So if you had to give, uh, other entrepreneurs one piece of advice, what would it be?
(47:29):
Follow your passion, try to find a problem that it really, that you are passionate about. Don’t do it for, don’t do it for glory. Don’t do it for the morning. Yeah. It’s at the beginning. They will not be any
(47:42):
Either of those things.
(47:44):
Yeah. Either of those things. So, uh, do it, do it because you’re really passionate about the problem, because that will keep you, uh, during the tough times that will be, uh, your guiding start your north of like, why you are doing this. It it’s always easier to answer. It’s like, because I really care or cause I care about my users and I really connect on this, on these solution. Don’t do it because it’s a big business opportunity because business opportunities are everywhere. Yeah. Or you can go and do something different. I dunno. Just sell NF NF.
(48:25):
Oh my gosh. It’s so true.
(48:26):
But that’s that some people might be, might be like, are really love FTS and Ft is my thing. So that’s great to go and do it and be the best yeah. That you
(48:35):
Can <laugh> yep. Yep. But follow your passion. I love that. Follow,
(48:39):
Follow your passion. So yeah, there’s always this anecdote of, uh, and I remember when I was doing, uh, I took some elective and art history and they were, were talking about Picasso and how the mother say, okay, what, uh, how he, she check him in many ways. Like you’re going to do something, just start to be the best. Yeah. You know, it’s uh, if you are going to be a priest, try to be the Pope, if you are going to <laugh>, you know,
(49:08):
And
(49:09):
If you want to be a painter, then strive to be their,
(49:11):
You the best. Yeah.
(49:13):
And so on. So, so that’s, uh, that’s probably have a step for perfection for, to try to, to be the best version that you can caring about the problem that you really care. Yeah. Um, and uh, I think that pieces will, will, will follow. At least for me is very gratifying. Every time that you talk like user. Yes. Okay. Yes. Yes. You are definitely better than, uh, the previous wear solutions and you are making a difference that by itself it justif all the growing pains of a company.
(49:47):
Yeah.
(49:47):
And every time I have to go and do taxes in different states or something like that.
(49:51):
Yeah. You’re like, I’m doing it for the farmers. I’m doing it for those avocado guys.
(49:58):
Exactly. These guys really care let’s to solve it. And <laugh>, if, if I don’t do it, I don’t know who else will
(50:06):
Do it. That’s right. No one else will, I’m doing it. Cuz it’s one
(50:10):
I been doing it. So it’s like, yeah, let’s let’s help that.
(50:13):
Good for you. Oh my gosh. I love that. Well, Carlos, thank you so much for taking the time today to talk with us. Thank you for sharing. I mean, your knowledge of this subject is so it it’s rich and it’s deep and you have shared it so willingly and I’m, I’m fascinated. And so thank you for being with me today and thank you for sharing that with the audience. This, this has been really, really fascinating and I really appreciate it.
(50:40):
Oh, thank you, Cynthia. Hey, very happy to, to be here to participate. Uh, yeah. Let me know when you want to keep having a conversation, we can go keeping machine learning models, applications or ethics, or <laugh> whether
(50:54):
To that point, if somebody wanted to find out more or they were interested, like, are there other like resources, books, podcasts, websites, anything that you’d recommend, whether on the entrepreneurial side or on the machine learning side or even in the, you know, micro climates and micro economies and that like any of that kind of stuff, anything you’d recommend if people wanna learn more.
(51:17):
Yes, of course. Um, great resources out there. Um, first of all, we, I will, I am biased that, of course I will recommend the book of my doctoral advisor, uh, William, she, uh, on machine learning applications on environmental science. Okay. Um, he’s writing the second version so soon will be released, uh, great book accessible with examples. Uh, it, uh, removes some of the needs around it, but also it tackles it from the perspective of physics of, uh, you know, uh, the part of the math behind it. Yeah. Uh, so you don’t apply it as a black box. I think that’s important. Yeah. To understand the machine learning limitations, because the problem of using a model that is out there is that if you don’t know what’s inside the hood,
(52:06):
You don’t know what you’re gonna get.
(52:07):
You’re not in control. Yeah.
(52:09):
They’re not in control
(52:10):
And you should be in control of your model. You should always understand at least if put result, like yeah, that makes sense. I, I made that conscious decision that it should, this output should be expected. Yeah. So that’s a, that’s on the academic side, on the, on the entrepreneurial side. I think that when we went into Techstar last year, there’s a very nice, uh, accelerator based in Boulder. Yep. It was beautiful to go there. Yep. Uh, love the Rockies. Um, we received a book on venture deals. Oh, Bradfield. Okay. Yep. I think that it gives a very interesting context for entrepreneurs that are looking for, uh, scaling their business and received, uh, support racial capital support. Yeah. There are different ways that you can be entrepreneurs. Some of them you could self fund. Some of them, uh, might chose the path to go and find support from different angel investors or institution investors. Yep. Uh, we had done a mix of everything from, uh, national science foundation grants, NASA grants. Uh, um, and uh, of course we have like constitutional investors. I think that this book is going to give a great context of, eh, how the relationships go when you have the conversations. Yeah. Eh, it’s a numbers game. You talk to as much people as you need. Don’t focus on only one. Yep.
(53:39):
And
(53:39):
That, you know that, and in many ways, time is money. You will have to spend your time. Yeah. Eh, speaking with potential investors, they are not all going to be a fit. Yeah. Eh, they also look for a fit in an investor because it’s almost a decision like having your best friend and, uh, a good investor can make or regular business in the future.
(53:58):
Oh. A hundred percent. A hundred percent. So,
(54:00):
Uh, yeah’s also, don’t don’t think that necessarily you are asking for money’s you are also offering them an opportunity to be part of something. Great. Yes. So it, this is not charity. This is a business opportunity. <laugh>
(54:13):
And uh, that’s not charity, you’re not begging, so now we’re not begging, that’s the thing. I always tell people, I’m like, why are you in this conversation? Like you’re begging for money. You have something for them too. It’s a relationship.
(54:24):
Exactly. It’s and, uh, is a relationship is like who you think that it will help you grow in the future? Yeah. To be the best version of yourself and the company. Yeah. 5g 10 years into the future. So be very strategic. Yeah. Um, because in the book, even they mentioned that it had seen cases in many cases of investors that can destroy companies.
(54:44):
Yeah. Oh, it makes my stomach hurt, so. Okay. Good. All right. Awesome. And so if, if, uh, our audience members would like to follow you, or maybe they wanna learn more about what you do or get in touch with you, what’s the best way for them to do that?
(55:00):
Uh, thank you. Yeah. They can go and visit us@benchmark.com and, uh, contact us is like a contact us form. <laugh> perfect. Or they feel free to, uh, email me at Carlos benchmark.
(55:12):
Perfect. Perfect. Happy to all. Awesome.
(55:16):
Hey, help the audience.
(55:17):
Awesome. I will include all that in the show notes so that everyone can get, can get that information really simply. Thank you again so much, Dr. Carlos Gaetan. I’m very, very happy to have had you here today. And I’m probably gonna ask you to come back maybe in a couple months and talk a little bit more about what you’ve done from an, from a business building standpoint, especially because there was some nuggets you just threw out about fundraising that I’m like, Ooh, I wanna talk about that. <laugh> <laugh>
(55:46):
Sounds good. Perfect.
(55:48):
Perfect. So thank you. Thank you. Thank you so much for being here today. We really, really appreciate you.
(55:54):
Thank you very much, Cynthia, and enjoy the conversation. This is fantastic. See you in a couple of months.
(55:59):
Perfect. All right. So thank you so much to my audience for joining us for this episode as always happy entrepreneur. And I will see all next time.
Thank you for listening to this episode of Precursa: The Startup Journey. If you have an idea for a startup and you want to explore the proven process of turning your idea into a viable business, check us out at precursa.com. Make sure to subscribe to this podcast wherever you listen to podcasts so you never miss an episode. Until next time…
(56:41):
Inspiration has struck. You’ve stumbled upon a great idea that you just know will change the world. So now what at Precursa, we provide the best tools to help founders and entrepreneurs. Just like you turn their great ideas into great companies that solve real problems for real people. We believe you are the change makers, the innovators, and the force that moves technology forward. All you need is an experienced guide to keep you on track and help you navigate the turbulent waters of starting up Precursa is that guide. And with us, your roadmap to successful launch is more direct with far fewer pitfalls, ready to change the world. B has your back.
Copyright © 2021 Precursa | All Rights Reserved | Site Created by Natalie Jark
Copyright © 2021 Precursa | All Rights Reserved | Site Created by Natalie Jark