UT08x04: Using AI for Sustainable Agriculture with Nature Fresh Farms and Solidigm

Much of the conversation on AI is focused on power consumption and chatbots, but there are so many other positive applications for this technology. This episode of utilizing Tech sponsored by Soy brings in Keith Bradley from Nature Fresh Farms to discuss how they're using AI to improve crop yields, plant performance, and make tasty tomatoes. Welcome to Utilizing Tech, the podcast about emerging technology from Tech Field Day part of the Futurum Group.

This season is presented by Solidi and focuses on AI at the edge and related technologies. I'm your host, Steven Foskett, organizer of the Tech Field Day events series, including Edge Field Day. Joining me from Soy is my co-host, Janice Roski.

Welcome to the show. Thank you Steven. It's great to be back.

Yeah, it's great to have soy presenting this season where we're bringing in lots of guests from different companies talking about different aspects of AI at the edge. And I'm particularly proud of this, not only because I know our guest, but also because you know, we're talking to an end user here. We're talking to somebody who's actually doing something productive with AI and not just applying a chat bot to things.

Couldn't agree more. I'm super excited about this. Super excited.

And you know, tomorrow is actually Earth Day, right? So this guest is very timely altogether. Indeed.

So let's introduce him. Uh, Keith Bradley is VP of IT over at Nature Fresh Farms. You know, we've seen Keith before at our Tech Field Day events.

We've learned a little bit more about na, what Nature Fresh is doing, and I think we're gonna learn more here coming up. Keith, welcome to the show. Hey, thanks Steven.

Thank you Janice, for letting me be a part of this. And again, it's always exciting to talk about ai. It's always exciting to talk about the edge and it's always exciting to talk about what Nature Fresh Farms does and how we do things.

And again, it's Earth Day, something that we take to heart producing vegetables for the world. So we love to do it. Absolutely.

I mean, you know, that that's, I think, gonna be maybe the right way to kick this thing off. Uh, a lot of people might be listing and, and maybe they have a negative impression of ai. Um, they think it's, oh, it's burning up the rainforests, you know, what could it possibly be doing on on Earth Day?

How could AI possibly improve our lives? But I, I'm gonna go out on a limb and say, you know, nature Fresh, you guys are actually doing some really positive stuff with ai. You're improving people's lives, you're improving people's health, and you're doing it in a way that benefits the earth.

Yeah, we're doing something I'd like to say that is amazing. It's always an amazing story of how we're looking to improve that taste of a pepper, tomato and cucumber for you. We're looking to find a way to get more yield from per meter squared, you know, and we're looking at how to integrate technology to make it part of that.

You know, we've been doing, growing for 25 years almost now, and AI is just becoming one of the biggest growers for us in how to improve things and how to push it right to the edge. Like, you know, IOT was a buzzword five, six years ago and I said, we've been doing I iot stuff for years and now we're doing the Edge and we've been doing that for years. So it's a great industry to be in and it's great to work with so many passionate individuals that we have at Nature Fresh.

So let's, let's talk a little bit about that, Keith. Um, so you, you talked a little bit about making these tomatoes and peppers taste better, right? Or blueberries or strawberries, right?

I kind of look like a blueberry today with my fluffy blueberry sweater on, but you, you know, when you talk about how you do that right, and, and doing things very differently in your farming environment, um, can you tell us a little bit about how you're collecting some of this data? How, how are you making these tomatoes taste physically better? So we look at everything that we can think of.

Everything under the sun we absorb how much sun the plant's getting, how much natural sun, how much artificial sun, how much irrigation the plant's getting, how much nutrients is taking in, how much is not taking in CO2 levels, anything like that that we can control or monitor we're taking in. And what we're doing is we're taking that data at the edge, analyzing it, and then helping the system make a decision to improve the yield and the taste of that product. 'cause what we wanna do is we want to increase the yield, we want to increase the taste and even at the end of the day, increase the shelf life for the customers out there.

'cause that is what is the most important thing out there for us. But all of this data gets collected, stored, ran through an AI engine, basically to decide how do we do it? Do we increase the temperature?

Do we decrease the temperature? Do we add humidity? Did the plant not get enough irrigation today?

Should it be adding more irrigation to this process, to the nutrient level go up? We look at all that and we analyze it every day and figure out how to do it better tomorrow based on the weather patterns out there, what's happening outside. Because we gotta remember, we control it.

You gotta think of our plants as they're basically a patient in a hospital room and we control everything in that plant's life from the day it's born to the day it's done growing and we control every aspect of it, what they eat, what they sleep, everything like that. And e exactly what you're talking about is really using AI to provide sort of a, a little robotic doctor or a little robotic farmer for each individual plant. And keeping an eye on them in a way that, you know, I think a lot of people think that farmers would like to do.

In fact, I think a lot of farmers would love to do this. They'd love to be able to watch all their plants and respond to their needs individually, but they can't because it's just not possible for, for, for people to do that. But AI with various sensors and and processing can actually do that.

You can keep an eye on every individual plant and cater to its individual needs and help it be the best little tomato it can. That's right. Yep.

And that's what we look for. And that's where we start to get into the conversation again of the edge. Because see, when we first started Nature Fresh, we started looking at an edge is a 32 acre range or even a, a a 64 acre range, very large section of greenhouse.

Now we're starting to push that technology not to be computed at the core, but at the edge now. So in smaller sections, more dedicated to just that plant because we might be growing three or four different varieties within one section of greenhouse. So we need to have a different algorithm, a different AI learning how to take care of that plant and, you know, talk about what you want to do.

Think about it. Most times when you think of things, most people are reactive. With ai we're able to become proactive and looking at how do we treat the plant the day before?

What's gonna do we need to do better with it? And it's looking at it and running simulations all night to say, this is how it is. So at our core, we run the simulations and then we, we through the edge, we uh, publish the right new, new large language model out to the edge.

And then that does the compute at the edge. So we're not having to have that tons of data. 'cause imagine we have hundreds of thousands of sensors out at the edge and trying to move all that data was just cumbersome on the network.

Yeah, I can imagine. I'm, I'm curious Keith, um, now that we're talking about the specifics of the devices that help you collect this data, um, you mentioned the sensors and everyone has a different idea fundamentally of like, what do we mean by edge. So educate us on what does the edge look like for you?

You've got the sensors you mentioned, but do you like have a physical edge server kind of sitting next to the plant that's collecting the data and then sending it back to something you have on-prem? Or can you tell us a little bit about your overall architecture? Yep.

So right now we've been really getting into, uh, into the native edge now and partnering with Dell to do it. But we're looking at the native edge and it's basically close to the greenhouse. So it's now basically not inside with the plants 'cause we're not quite there.

What we do have some out there, but we're getting 'em right close in a skewed room right on site that does all the local data collection. It collects all the data from there, stores it, and then it kind of condenses it uploads it overnight to the core servers, which is anywhere from, uh, you know, a hundred miles away to 10 miles. And then that core, uh, core server updates, the model updates, how it would look at the reactions, and then republishes that out for the next day to start.

So the new, the new next day starts with a new model update a process and what it should do with it. And that way we can refresh and update things as they've happened. Plants again are something, we learn something new every day and that's what we're looking at.

And we take old data to make it good And you really are tailoring it to the needs. I don't know, is it really individual plants? Is it, is it groups of plants?

I mean, is it It we like to call it, I guess you can call it, we, we, we call it um, we call it rows of plants. So we, we customize it to each row or each section of plants. 'cause we can't do each individual plant, but we can kind of customize it to that section.

'cause each section does it out. You know, we call 'em houses 'cause the, you know, it's got the nice peak like a greenhouse, you know, the houses and rows that are how we kind of divide it out. But within each row we have so many plants.

So we know, hey, there's, you know, 3000 plants in this row. We know how much they're gonna get. We divide it, we know, okay, this is how much is going on.

And then we also have cameras and even people taking pictures of stuff and uploading it to say, this is what the plant looks like today, this is how it is growing today and what it's looking like. So we're constantly updating what it actually looks like in the greenhouse and then going into what it is. Then the next thing happens is that edge, once we pick that pepper pick that cucumber now moves for me.

It moves, it moves now to the packing line and that's where another spot where you have to do ana analytics in real time because that pepper goes through that line and oh my gosh, it feels like 30 seconds, but it's about four or five minutes and it goes through and you have to make decisions. Is it a number one? Is it a number two?

Is it a number three is the color variation rate. And we've just started to play with using AI to identify that PLU on each tomato to verify is that PLU on there. And if it's not, we literally have about a second and a half window to box kick that box off.

So that edge is now the packing line and then you go again to the next edge. And the next edge is actually being in storage when we're getting ready for the trucks to go out. And we kind of commodity that right to the end when it gets to the grocer.

So it's, it's an amazing edge and it's something we keep doing, but we're keep moving technology in our stack closer to that edge each time now. Now we got ones where the servers on the packing lines, you know, five years ago we didn't have that. We had one core server, but we're trying to move that compute closer because of that time differential.

And you know, you mentioned the time differential being important and having the sensors and you know, collecting all this data, you know, nightly and then taking a look at every plant. Um, you, I think you mentioned, is it it is or it is not a very, um, uh, you know, data intensive process in terms of how much storage you need. Do you, do you rely on high capacity storage or Oh yeah, yeah.

We, we store tons of data. Uh, we collect, so each plant averages out to, it's a, it's anywhere from about four to five megabytes a week that they produce per plant. And at any point in time we can have 3 million, about 3 million plants within the greenhouse.

A lot of it's transactional stored for anywhere from seconds to minutes. But all that data is hitting some type of storage device to be analyzed and then either dis discarded or stored or long-term stored. Um, that's a great thing about what we've done done is we've now have dates and history going back, oh geez, almost 20 years in some locations of what they've done.

Crops might have changed, but the weather patterns and what they do is what the growers sometimes look at too. Are you storing the stuff on, on hard drives, solid state drives? What's your drive of choice?

I'm always been a solid state. Um, I have to admit this, I go back, I bought my first solid state server for Nature Fresh 21 years ago. Like we were just, they were brand new just out there and I said, this is the way to go.

And I remember 'cause it was such a hard conversation to sell because they said, oh, that solid state drive won't last five years for that server. It won't last. It's just a fad.

And the compute power that we got outta that changed how we worked. And from there I fell in love and it never disappeared from the farm. Everything from laptops that do a little bit of AI workload on them to anything has a solid state.

And it's been that way for all, I'm gonna say about 15 years now. So I'd like to say that was one of those things that I fell in love very early because of that processing power and the longevity that you do end up getting out of them. Yeah, that's I think one of those things that, uh, people have found.

I mean certainly many of us have found that with laptops, right? I mean you have a laptop with a hard drive versus an SSD, it, it's transformative. You wouldn't think it was 'cause that's, you know, just the storage, but it's so much better that it enables you to do a lot more with it.

And I think that that's what people are finding as well in um, edge servers and uh, AI servers. I can't imagine somebody imagine trying to build an AI system around hard disks. I mean, uh, yeah, maybe there's hard disks storing some archival data somewhere.

But if you're gonna process it, you're gonna process it off SSD. Yep. When we switched to an all flash storage for our information like that, we noticed again, uh, 2020 was the year we switched to all flash and for our, our active state of information.

And it was a life changer for us. We literally went from hours of compute down to minutes in some areas and everything started to become, so it was that first day that that really started to compute that we can do this in real time, that we can actually start to make this happen. Once we started to really use solid states in that way, like everything was on solid state, it was not okay.

CarVal nodes that we have now. We stole our, our our security video, stuff like that on the spinning drives. Let that sit there 'cause somebody can wait a few seconds.

But transactional data that's making a decision has to be on that flash drive. It has to be there and it has to be accessible right away with the right bandwidth attached to it. Yeah.

Especially if you're running the AI workloads that you're running. Right. They just, the other stuff just can't simply keep up.

Um, I have a, a couple more questions. This is more on the overall sustainability side of things, right? There's a lot of hot topics in farming right now, particularly around soil prevention and carbon and um, how is your farm more sustainable than say, a more traditional farm?

Or is it, So I'd really love to say talk about our sustainable story and it is a, it is a great story. It's, it's one of the passions we have at Nature Fresh is our sustainability story. You know, we'll, we'll start at the simple stuff.

The water. So when we irrigate the plant, whatever the plant doesn't absorb, we take back and put into the plant next time. So we, we reclaim water every moment that we can.

So nothing goes to waste. We don't just irrigate a whole field, we just irrigate and then take the water back. So we're a hundred percent controlling our water.

We're recycling rain water for stuff like that. Now you get into the sustainability, we don't really use soil for a lot of our plants. We use ground up coconut husk.

So our growing medium isn't even soil because we want that neutral non nutrients non-growing state for the plant where we know exactly what we give each day. So we're not even touching the soil. All of our plants are actually elevated off the ground.

They're on a tray, they're there because it's multiple things. It makes it easier to work on when the plant's in the air, you know, think what the field to tomatoes, they're all down the ground. You're working on it.

It's right there in front, right at eye level and you work on it and you look at it. So you look at stuff like that, then you go into even how we control our energy. You know, we have a large amount of natural gas that we use to produce heat with that production of heat creates CO2.

We actually take that and give it to the plants. 'cause that's your natural stimulant. And then they give out O2 after they absorb it.

So every part of our process, we are doing something to keep it sustainable then. And I love I, if you ever get a chance to come by, it's always a great passion of mine. Our one facility, we actually burn wood chips to eat the greenhouse Taco Bell.

The best way to heat the heat a greenhouse is just good old natural wood chips that would've hit the landfill. We're taking all that, we're burning 'em, it's gone. And it, it's such a clean way of doing things, you know, um, all of our cardboards recyclable.

We, we control every little bit of the plant and even the way we try to, we, we control pesticides too is that we don't really use, we don't use pesticides in the same manner. We don't use it in the same way because we are all controlled environments so we don't have to worry about it. Uh, it's, that's fascinating.

And of course all of this is incredibly compute and data intensive Right. To be able to do any of this. Yeah.

Even going back to where, where it really started to is I go back to even the days when, why did we start to track all this? Part of it was even for, for food traceability is we wanted to know where this pepper came from. So we created a program so we can say this pepper, it came from row A in greenhouse B picked by this person two days ago.

So it's just something that was amazing to do. And the data that we started to get things allowed us to realize, wow, we can now make real time choices with this data. It's, it's simply amazing.

And, and being able, I mean just tying this all back to, like I mentioned, earth Day, right? Everything you're doing Yeah. Um, to improve the planet with what you're growing here, right.

Um, is the crux of it is, you know, your innovation that you're doing, but also having, uh, just really good data at your fingertips and being able to utilize all of that. It's just, yeah, I wanna come to the farm and visit. Yeah.

It's, it's an amazing thing to see and it's a passionate thing that we do. And again, that passion that we have to grow a better tomato cucumber is what makes us strong. The sustainability of what we do, us being able to hold onto that and being able to hold on how we control that greenhouse and get that bell pepper to you better, faster, quicker is one of the major things that we can do.

One of the interesting things about uh, AI at the edge is getting back to one of the things you said earlier, Keith, that you can collect more data than you normally would be. Because if, if you have to centralize your data, if you have to ship it all back to a cloud server or a data center somewhere, then you just can't collect as much data. Whereas if you can process it closer to the edge, closer to the sensors, you can collect a lot more.

Would you say that that's sort of the, the secret of what you're doing with AI at the edge? I'd like to say that's one of the secrets out there. That's one of the major things that changes how we do things is that data is right there at the edge.

It's able to make that decision. And we've noticed it more and more as we push the technology from that core server, from that cloud server right to the edge. And that again, that's, that's what's, that's that, yeah.

I like to say it's one of the secret sauces, but it's definitely one of the things that's helped us allow to give our growers and our grower scientists that secret sauce to continue to improve that plant. And can you tell us a little bit, Keith, about what that server kind of looks like? Again, I'm, I like to learn more about the architecture, right?

But particular type of server. Are you, are you running on? So we're really based on, on a Dell VxRail.

Um, we, we really have that, it's a nice combination of Intel and a couple of Nvidia uh, a thirties and a forties cards doing the TPU compute on them. But we've recently really just upgraded now to the Dell Native Edge. So it's actually a, a, a smaller compute server, but it's designed to have a blueprint that you can push out to it.

And that's really changed the way we've done things in the last I'd like to see even eight months now. We've really seen how we can do it. 'cause not only can it help us control the greenhouse, but we're a accessing the packing lines, accessing more PLCs out there and collecting not only data about the greenhouse, but data about the packing line that we never seen before.

Never really thought before that we would get. And it's given us a new insight on packing and how we do things and again, making that reaction happen in real time. So that structure is what we do.

And then there's, it's a nice way of doing things because as we add more native edges, we can deploy a per blueprint out that's saying, this is what we think you're gonna do. And then we can adjust that blueprint to be specific for a tomato, even specific for a specific tomato, um, variety. 'cause we have multiple varieties of tomatoes, peppers, and cucumbers.

So it might look like a yellow pepper to me 10 year, 15 years ago. But there's actually subtle differences between each one. So now we're looking at dividing those compute out between each individual crop and each individual variety.

So it, there's just that little bit of difference. 'cause every variety and every plant needs a little, uh, TLC and a little love. And that's where that data helps us.

And you're actually running these servers you said, um, really close to the line. How close to the line? Are they like, like sitting underneath the, underneath the conveyor belts?

Are they uh, Yep. So in the packing line now? Yep.

We had one, uh, when we first started. Yep. It's like right underneath the packing line.

It's actually right there next to it. Uh, in the greenhouse. You know, we got them basically as close as we can to the crop that's not gonna get wet hot and stuff like that.

We've really been looking at how do we put stuff right into the crop now. And the Native Edge does have that platform where we're looking, okay, can we actually start to help build even the robots to go down the road that has that and have that connectivity and that scalability. 'cause again, you gotta think we have to build a scale.

You know, we are, we're 250 acres of greenhouse greenhouses are gonna take over more and more acreage across the world, and we're gonna get bigger and bigger. How do we scale that out? Because it's just getting so much more every day and you gotta have a great solution to scale.

Things like that. You know, it's funny, uh, for years I've been talking about, I'm making the joke that at the edge, like the server is underneath the fryers. I'm gonna change that.

Yeah. I'm gonna start saying that the server could be underneath the tomatoes and, uh, and peppers from now on. Yeah.

Because, uh, you know, it can be, right? Yep. It can be.

And actually when we did our, our proof of concept on it, when we originally did it, basically I just plugged in a network cable to the network for the packing line network cable to give it access to our, our network and threw it underneath the packing line and came back a week later and had to dig the, uh, debris and the, the, the little bit of garbage that falls from, you know, anything like that to clean it off, dust it off. But it worked right through it. And, and you know, that's what we need to have and that's what it is.

You know, some days we're, it's funny 'cause we're a very clean environment for the plant because there is no dirt and stuff like that, but there's still just debris, leaves, everything else that comes through it that it's like, wow, we forget how dirty it can get sometimes. So, so you mentioned, you know, the scalability piece of it, which I think is really interesting. Are you guys based like on a farm or you inner city or what is it?

What is your Well, so we have two farms. We're located in Ontario and Delta, Ohio. Mm-hmm.

And it, where I, we come from actually was a very, uh, very large farming land as it was, but we only could farm, you know, for three months of the year. Like basically we were very centralized around tomatoes, but our tomato season lasted about four weeks. So when greenhouses started to come in, we started to change that from being four weeks to being 10 months of the year.

Now with artificial lights, we're now producing tomatoes 12 months of the year. Like every day we are picking something and doing something, and it just grew and expanded. And so yeah, we're, we're kind of really put a greenhouse anywhere.

We do t pick typically pick fields, but a lot of times we pick the fields that you don't think of. Everybody's looking for that lush farmland. We not as worried about that.

We more need the infrastructure of a small city in some ways for power and for water and for things like that to help us do things. Yeah, that's really interesting because, you know, here in in Akron they're reusing some of the old industrial plants for vertical farming and greenhouses and you know, I think that, you know, driving through the city, you might see an old brick factory and think, well, that's not, you know, that, that that's just an old factory, but actually they're growing in this case, for example, microgreens in vertical farming in, in that, uh, in that plant. Did, do you envision, uh, growing, uh, peppers and tomatoes and cucumbers, uh, in, in environments like that too?

Yeah, one day I really do think that we will be, um, you know, when we really get a good understanding of how light works and how the plant's absorb even better than where we are today, um, we'll really be able to artificially give them everything they need. Uh, we still do love the sun. It's, it's our best friend, you know, I'm actually, you know, uh, out here and it's a beautiful sunny day, so it's a great day for the plants.

All that energy coming in is, is a great thing for 'em. Artificial lights definitely do help. They change how we do things.

They allow us to do the winters, but there's nothing better than the sun. And give us another five, 10 years. I definitely say that, that warehouse with no light, we'll find a way to make that work.

And that reminds me, one of the things that they mentioned as well was, um, you know, the lack of pesticides that they need because it's completely in enclosed, so they don't need to use a bunch of artis, you know, chemicals and so on to keep the, keep pests away. There just aren't any pests. Yep.

And we do, we actually go through a complete clean out, uh, pre oil every year, depending on the variety, depending on the type of, uh, peppers that we have or cucumbers or whatever we need to do. We'll strip out the greenhouse, get rid of every pest we have, start with the freshest day, zero pests, no nothing there, and then we start to grow. And that gives us that ability to have that perfectly clean tomato, perfectly clean.

It's no pest. There's no nothing to bug the growers. It's just a matter of focusing on how do I make that plant better?

And again, touching on the AI and, you know, we're partnering with many people, people that, that are doing ai. How do you do this stuff to make this grow better? Because there's no one, there's no one formula that's gonna make it perfect.

So we just keep doing more and more. One of the last questions I wanted to ask you, Keith, is, and I'm sure the audience is curious, we talk about the very beginning about, um, the data really collecting, you know, how much better can we make this tomato taste right? But is there, is there any way to collect to make sure that that tomato or that pepper has the same nutrient value that it might have if it was grown elsewhere?

Do your products have the same nutrient dense value? Yeah, we, we do all kind of nutritional tastes and we have a, uh, one of the honorary things you can get if you worked at Nature Rush is we actually even have our tasting club. So you actually get to taste in the many different varieties we have.

I've never done it, don't really know if I have the palate taste to be able to do it. 'cause you gotta pass it. So I don't really wanna go to it and fail it, but we really do watch what it is and we make sure that it hits the nutritional standards that you wanna see and that we match things.

And, you know, that's just what we do and we get the passion we have. That's so cool. I I I, I would totally, even if I failed, I would totally want to join the tasting club too, because you get to taste all this delicious produce all the time.

I'm, I'm, I'm loving it. Wow. Yeah.

Yeah. You get to go into a little dark room, have three or four different cherry tomatoes in front of you, strawberries, peppers, cucumbers, and saying, which one do you like better? The, and you know, we actually, it's a special light so you can't tell the look of 'em.

So it's, it's totally clean and, and all that. So it's, it's a really neat process. That sounds just great.

What do you think, Janice? Uh, you wanna join the tasting club with me? Uh, have some tomatoes.

I I think we're gonna have a field trip here very soon. Yeah. Well, you guys are welcome.

Wow. Just incredible. And, and you know, the coolest thing, Keith, is that, again, this is not the same, you know, let's, let's make another chat bot.

This is not, you know, oh, I'm gonna make an expert chat bot that the farmer can talk to, to learn about tomatoes. No, this is directly influencing the, the growth of healthy, wholesome, pesticide free. You know, I mean all, all, all this.

I mean, it's, it's incredible to think that AI does that too, and that edge computing makes that possible and, uh, storage makes that possible. Uh, I really appreciate you coming on here. Thanks so much.

Yeah. Every time. Thank you so much for having me.

I always enjoyed talking and sharing our story and sharing the t and for technology that I have and the passion Nature Fresh has for what we do every day to grow the peppers and tomatoes. Well, thank you so much for joining us here on this episode of, uh, utilizing Tech. As we wrap up this conversation, Keith, where can people learn more about Nature Fresh, uh, generally and about what you're working on specifically?

Yeah. com. That's our main website.

Uh, you can look me up on LinkedIn. Uh, you know, I, I do share some stories here and there of what we are doing. And you can find Nature fresh on Instagram and Facebook and all that fun stuff.

How about you, Janice? Um, where can we catch the latest from Soy This Earth Day? com/ai, uh, to learn more about what we talked about here today.

And then, uh, stay tuned for the future because you never know. We may have another storyline, uh, to share with you all. Excellent.

Thank you so much. And, uh, thank you everyone for listening to the Utilizing Tech podcast series. You'll find this podcast in your favorite application as well as on YouTube.

If you enjoyed the discussion, please do consider leading us a rating and, and a review. We'd love to hear from you. This podcast was brought to you by Soy and by Tech Field Day, part of the Futurum Group.

com or find us on X Twitter, blue sky, and mastodon at utilizing Tech. Thanks for listening, and we will catch you next week.