Inside Ukraine's Azov Drone R&D: The Engineer Building AI Weapons 18 km From the Front Line | Alexander Palamarchuk

The most consequential arms race in the world right now isn't nuclear, it's software. Craig Smith speaks with Alexander Palamarchuk, an engineer in the R&D department of Ukraine's Azov Brigade, calling in from approximately 18 kilometers from the front line in the Pokrovsk region. What emerges is one of the most technically candid accounts available of what drone warfare actually looks like from the inside: how Ukraine went from homemade reconnaissance drones in 2014 to AI-guided systems being developed and tested in real combat today, how the jamming arms race has forced his unit to develop custom frequency systems spanning 100 to 3,000 megahertz in a constant search for clean windows Russia hasn't yet closed, and why the tank - once the defining weapon of land warfare - has been reduced on the modern battlefield to a mobile jamming platform. The most important distinction Alexander draws is one that rarely surfaces in mainstream coverage: AI already exists that can recognize and classify vehicles and people with high accuracy. The unsolved problem isn't recognition, it's discrimination, determining with certainty whether a recognized target is military or civilian. That gap is the only thing standing between today's AI-assisted drones and fully autonomous lethal systems, and Alexander puts the timeline for closing it at approximately two years. The US maintains an official policy of not developing fully autonomous lethal weapons for ethical reasons. On the battlefield 18 kilometers from where Alexander is speaking, that policy is being outpaced in real time, and he is clear that NATO's software advantage positions Western countries to win that race before anyone else gets there. Subscribe to Eye on A. I. for weekly conversations with the people building and deploying the future of AI.

• 0:00Speaker 00

  The US has said that will not develop fully autonomous lethal weapons for ethical reasons, but it seems on the battlefield, every side is trying to get an advantage. But in two years, you'll be able to field fully autonomous drones. I mean, that's a frightening world. I would guess that that technology is going to be on the battlefield very soon, if it's not there already.

• 0:26Speaker 01

  Drone with AI is on the battleground, China, their presence. There is some pilot AI, they are good at recognizing targets, but it doesn't recognize, for example, right target or wrong. It's very good at recognizing, but it is not good enough to know is it enemy or not. There's a problem. The network countries are trying to get the right weapon, I mean, right model. So they are trying to do studies this way, or actually.

• 0:55Speaker 00

  Paradise is the callsign of a man named Alexander, a Ukrainian drone specialist serving with the Azov Corps of the Ukrainian National Guard. Azov began in 2014 as a volunteer battalion. It drew international notoriety for its links to far-right movements and was later incorporated into Ukraine's National Guard, where it has grown in one of the country's most prominent combat formations. Paradise joined in 2023 and now works in an R and D unit, developing drone technology near the front.

• 1:36Speaker 00

  I spoke with him remotely about jamming, autonomous navigation, AI target recognition, and how drones are rapidly reshaping the battlefield and the future of warfare. First of all, I wanted to ask you about the history of Azov. When you joined, when Azov started using technology?

• 1:59Speaker 01

  I was joined to Azov in 2023. And work only with drones, is the different kinds of drones. If you're talking about Azov, as I know, the best usage of drones in Azov was like in 2017. They was used to reconnaissance firstly and then to strike some bombs.

• 2:32Speaker 00

  To strike like drones as anti-drone warfare, is that what you mean?

• 2:38Speaker 01

  They was used to hit different kinds of targets. As a first of all, it was used as reconnaissance drones. So you can just be in some kind of drone and using when you can even see your enemy with your own eyes, but it gives you a good angle to view what is going on on the enemy side.

• 3:07Speaker 00

  Right. But the war didn't start, the invasion didn't happen until 2022. So in 2017, where were you using this?

• 3:19Speaker 01

  Actually, it started in 2014 for us. Actually, even only two people can find a video with Russian regular army in Donetsk region. So it was used with the same enemy from 2014.

• 3:37Speaker 00

  I see, yeah, in Donetsk.

• 3:40Speaker 01

  That was started with home-made drones. Then when some kind of DJI Fandoms arrived, guys tried to use them. And I know they were glad that it gives good possibility for reconnaissance.

• 4:00Speaker 00

  Okay. And when you joined, they were still using DJI or home-built or when you joined the effort?

• 4:12Speaker 01

  Even for now, we are using some Mavic's and Matrice drones for reconnaissance, for some kind of bombers with light bombs. Mainly, we use them as bomber. And now, on this year, we use them mostly for reconnaissance, because for now, we have more possibilities and to take bigger bombs or fly longer, staying there for longer, actually.

• 4:50Speaker 01

  They have good cameras now, and we are on the way to change the Chinese drones for our own-made drones, I mean, Ukraine-made drones, or our partners-made drones. We're on the way for this.

• 5:08Speaker 00

  One of the big problems with drones on both sides is jamming. And I've seen all of these different jamming systems, but also systems to evade Russian jamming. I mean, how do you handle that? I know there's the fiber optic line that unspools up to 40 kilometers or something, but there's, I've also seen drones with Starlinks mounted on them. So there's satellite communications, which I understand is unjammable.

• 5:51Speaker 00

  Is that, can you talk about the jamming issue on both sides?

• 5:55Speaker 01

  Yeah, I can. I will divide it to three parts, like radio, like fiber optic, and like internet connection. Let's just start it with radio. This is a classic drone switch I started to use in 2023. First, we use its common frequencies, the same as LTE, for example, or some 2.4 gigahertz. It was regular. Then we meet some jamming from enemy side. So we are trying to use the possibilities which we have had with ERS.

• 6:40Speaker 01

  And we can move from 700 megahertz up to 1,000 megahertz. And in this frequency range, we're trying to find the possibilities for us to fly over the enemy side. And actually, until I guess the need of 2024, it was more than enough. Then we moved to our ERS, maybe you know, kind of ERS, which are made in Ukraine. This is a part of, this is based on common ERS, but it's made for our needs, like to

• 7:23Speaker 01

  have a stronger communication with the drone and the ground station, which was helpful to fight the jamming. For 2025 year, we moved to not regular ERS receivers, transmitters. For now, we can use almost from 100 megahertz up to 3,000 megahertz. And we are trying to find some frequency ranges, where can we work freely. And it mostly depends on where you're trying to fly.

• 8:07Speaker 01

  Just for example, we couldn't fly to Crimea region on common frequencies, but a guy who stands five kilometers next to us could freely fly on these frequencies on their point of interest and for good for them. So we moved to, we are trying to resolve our problems and try to find the solution for this by ourselves in the unit.

• 8:45Speaker 01

  For now, we are using not ERS for radio communication, but some kind of Ukrainian stuff, or our partner stuff.

• 8:58Speaker 00

  Okay. And then what about internet communications via satellite?

• 9:03Speaker 01

  Actually, maybe you know some names of drones that use it even from 2022 or 2023 year, like Mazzes or something like that. This is, this is a common internet connection. Actually, we are Starlink or some satellites like this.

• 9:27Speaker 01

  That's all, actually, what is, there is nothing, there is no, actually, secret from this, you can use Marlink, which can be routed via internet or via some local network or something like this. And you have your control and you have a video from a drone, so it gives you possibilities to fly where the something else wouldn't, wouldn't fly. You will have actually no problems to fly over some kind of buildings or some, something

• 10:10Speaker 01

  like that, even if you have a phone from your ground stage.

• 10:13Speaker 00

  Is there a possibility to jam the Starlink connection?

• 10:17Speaker 01

  I guess that is possible, so we never actually, is that what, what have I, you're on this? So it is possible. I have it on battleground for now, but I think that it's possible somehow. It is not, this is not the thing where you could find easily, but you need some kind of very expensive stuff, which could easily be hit by something else.

• 10:52Speaker 00

  Are the Russians also using something like Starlink?

• 10:56Speaker 01

  Yeah, yeah, yeah. You can find even videos where they're using Starlink.

• 11:03Speaker 00

  Right. And so at this point, neither side has an advantage, is that right? With technology?

• 11:11Speaker 01

  This is the kind of technology or nation that we could find the solution in our own unit and by one person. And when you are trying to solve some problem, like you are jammed or something else, you can find actually even in three parts of internet, very, very, very many. Very much information about what you can do with it. And this is good, where we have some kind of engineers who can sources.

• 11:53Speaker 01

  So we are good at making some very interesting projects, very interesting solutions to some kind of jamming, if you are talking about jamming. But I guess Russians are good in to do it more, I guess, in kinds of, as a part of system. They are good at producing many, many, many drones and this of using this kind of communication, for example.

• 12:31Speaker 00

  Yeah. Are you, are you guys using any artificial intelligence in your?

• 12:37Speaker 01

  Yeah, using as early programmers to make our lives with code is much, much easier. Example, actually, when you're using, when you're working with some data, example, it is very helpful too.

• 12:57Speaker 00

  Yeah.

• 12:58Speaker 01

  It's, it's a common, I guess, is a way to use AI. Mainly use it in, actually, the same way.

• 13:10Speaker 00

  Yeah. And that's, are you using cloud code or what, what model are you using for, for?

• 13:19Speaker 01

  Actually, when they have some, something we should code, we are trying in this model or this model or this model, and which would good, which would get us better results, we use that. So I couldn't tell that this one model is great for us, but they use different others which works well, we use it.

• 13:44Speaker 00

  Yeah. And are you using AI in drone, to help the drone navigate or in to auto?

• 13:56Speaker 01

  Yes, of course.

• 13:57Speaker 00

  Yeah. How, how you're using?

• 14:00Speaker 01

  There are two kinds of auto recognition for drones, like where you track some pixels or when you recognize some kind of UAVs or something like that. Or not UAVs, but some kind of cars, was this. When you're tracking the pixels, you move some little, for example, forward, and you just lose your target and the drone couldn't recognize this car, this is armored vehicle or this is civilian cars. It's not the great, but because you always, the operator should track what the drone

• 14:44Speaker 01

  is doing, you could, you should, you must, actually, you couldn't use drone as like, move to this area and hit everything you see, because you need to recognize this civilian car or this, they already destroyed car or some, all this. And we are, if you are talking about AI models for auto recognition, they are much clever. They are much more clever, they can recognize by own drone, is this, is this correct target for hit or this isn't.

• 15:27Speaker 01

  But operator should be present, and he has a drone control, just in the case that this should be anyway.

• 15:39Speaker 00

  So you're saying that they're, they're models for, for recognizing objects? And what about navigation? I mean, how autonomous are the drones today? You know, there's FPV, of course, where the operator is watching through a screen, either goggles or on a screen and guiding the drone that way.

• 16:15Speaker 00

  But if you have object recognition in the drone, can you do, can you, can you make it autonomous so that once it locks on an object, you no longer have to guide it?

• 16:32Speaker 01

  This is the case where the AI is help, because it put track the target correctly. If you are talking about navigation, I guess you know that GNS is almost everywhere is jammed and near the battleground. So we are, we have two ways to resolve this problem, that drone couldn't navigate by itself because of the GNS jamming. This one is visual position, as a drone should have a camera on the bottom of it.

• 17:11Speaker 01

  See, it's, check what is above this and compares to maps, which it preloaded to the drone and the result also. I'm here. So it can move to a position with some kind of accuracy, with some progress, accuracy up to fifty meters, for example. Difference, visual position gives different results.

• 17:52Speaker 01

  And the other, other way to solve these problems, to use beacons, like if someone communicates with ground station with one, two, three, beacons, it gives you a point where it is, depends on the time which it communicates with beacon, with the ground station, compares the time it need to communicate with ground station.

• 18:32Speaker 01

  Beacons, it gives a different time, so you know, you are closer to ground station and more and far from beacon. You know, this time needs to, to communicate. For, for example, you need to 10 milliseconds for one kilometer, for example. If you have 60 milliseconds, you are far as about six kilometers. And beacons help it to evaluate where you are.

• 19:08Speaker 00

  I see, yeah, uh, and uh, have you guys experimented with fully autonomous drones?

• 19:17Speaker 01

  Uh, yep, yep.

• 19:20Speaker 00

  You do, can you tell me how that works?

• 19:23Speaker 01

  Uh, actually, you just place one on the ground to, uh, then you, um, make some kind of function for it. Uh, it's a take off by itself, so that calibrate, for example, compass or maps or find it, uh,

• 19:46Speaker 01

  radio link with beacons, some kind of this, and then move to the place or area where it should be. Um, then, uh, that's the case, where is AI needed, uh, cause the problem is that you should mark that it's correct target. Um, just for now, it is impossible to, uh, hit only enemy targets, so it would be civilian, so you couldn't hit this kind of targets, this civilian car, even if enemy uses civilian cars as its own. So you, this, we couldn't hit this kind of target, so you need

• 20:30Speaker 01

  to be sure that it's enemy one hundred percent, uh, that's operator need. Um, um, if it is not a coming concept drone, uh, it is possible to make self autonomous solution, so just move to this area, watch this, watch this, watch this, and then it returns back.

• 20:54Speaker 00

  Are the Russians using fully autonomous, from what you can see?

• 20:59Speaker 01

  Um, I guess, we know about hits, uh, it's using GNSS, some kind of CP antennas, and the, it is, uh, uh, many jammers to jam CP antenna, so up to,

• 21:20Speaker 01

  I guess, they have some problems with this for now, uh, cause we saw, uh, CP antennas with sixteen elements, so you need sixteen jammers to jam this. Um, they moved to Starlink, for now they're using it, so if you saw the one of the was, they hit that train, civilian train which most, uh, hockey, uh, hockey, uh, if, if hockey first, I guess, up to. Even, even now, you could find some, uh, those fully automated, they just, um,

• 22:00Speaker 01

  pick over this and move to this area, hit, hit this, uh, and if it would be jumped, it hits anyway it could, I don't know, uh, if the drone thinks he's there, uh, to hit this, or even if it's jammed.

• 22:19Speaker 00

  Yeah, I mean, I mean, because the Russians don't seem worried about civilian casualties on the Ukrainian side, uh, so I'm, I'm wondering if they're gonna move to fully autonomous that will target any, uh, any vehicle or any, uh, uh, person in the, in its range.

• 22:47Speaker 01

  Actually, it's, uh, it's much easier for them, they don't care about any civilians or something like that, if you could hit any target that can, that move faster, for example, than twenty kilometers, and car, vehicle, uh, this is a target. It's much easier than if you can just, uh, move to this area, hit everything that is moving. So, it's much easier than hitting, but, uh, they are too much, uh, too many civilians was hit it like this, uh, actually, they even have

• 23:33Speaker 01

  telecom channels where they show that, oh, this was, uh, a woman with, for example, bicycle hit it. Even in this medic, in manual control, they saw what they did, it's okay for them. So, it's much easier for them to make some absolutely, uh, uh, autonomous missions, cause they shouldn't care about the type of, let's hear that, hit it.

• 24:05Speaker 00

  I mean, do you analyze the remains to see what kind of, I don't know if there would be anything left, but to see how autonomous the Russian drones have become, I mean, do you periodically uh, capture one before it's, uh, detonated and, and look at how it's being guided?

• 24:29Speaker 01

  Um, yes, of course, of course, we are trying to, um, and get something interesting from their engineers. I absolutely, uh, confident they, uh, they here, let's, uh, do something like we do. Uh, they are trying to, uh, reproduce our good ideas, we are trying to reproduce their good ideas, and the, the, the device that should work. Right, um, cause if you don't honor to use good idea of your enemy, you lose some kind of Yeah.

• 25:10Speaker 01

  Of good, you losing some good ideas, uh, uh, they have, uh, this Shahed is very, uh,

• 25:20Speaker 01

  efficient, actually, they can produce a lot, uh, and hit, this is, uh, uh, this kind of drones, uh, isn't so accurate for now, but they're using some mesh systems, some Starlinks, uh, to make it, uh, uh, to have a control of it, to get a video from them, to get some kind of telemetry from them, uh.

• 25:50Speaker 01

  Actually, everyone wanted to have a fully autonomous drone, just press X to win, just press X, ten thousand of drones fly, hit every enemy, uh, but for now, I see that both sides of this conflict of, of trying to get as much control drone as it, as it was. That we are, uh, everyone's trying to using some, uh, network systems, uh, like mesh or Starlink, so any kind of, uh, internet connection to the drone.

• 26:29Speaker 00

  Yeah, I, I mean, what you said about launching, you know, ten thousand drones at once, uh, do you think if the war continues, it'll get to that point? Uh, that, that Russia could launch ten thousand autonomous drones? Uh, so, so the drones, uh, kill everything or, or hit everything that's moving? Uh, I mean, that's pretty frightening, uh, do you think?

• 26:57Speaker 01

  I don't know that Russia is closer to this than we are, actually, uh, yeah.

• 27:07Speaker 00

  Yeah, except that, except that, as you said, you're, you're fighting over Ukrainian territory, and you don't wanna hit Ukrainian civilians.

• 27:17Speaker 01

  Actually, you know, we have, uh, the Ukrainian army, we have an invasion in, uh, uh, Russia, and the same, the same thing. The civilian, they couldn't hit it, but we saw this, this, uh, battalion commander here of Russian army, no, we couldn't. We couldn't, just civilians. And, uh, the partners looks what we are doing, they saw that try not to, to hit civilians uh, as possible. So, is, is that's one of the problems to use, uh, sort of autonomous, uh, coming as a drone for us?

• 28:01Speaker 01

  Uh, what is it possible to make some kind of autonomous, uh, or, or, or missions? Uh, when you couldn't hit someone? Uh, is it, uh, do we are close to, uh, the life, ten thousand drones, for, for example, ten thousand drones?

• 28:26Speaker 01

  Actually, yes, I guess, the two, three years, I mean, I have the situation where we launch many, many, many drones. The one problem is, uh, you said correctly, the navigation of the drone. Um, if you have some kind of systems that could, uh, um, and recognize the enemy correctly, uh, without hitting any wrong targets, it's possible to use, uh, many, many autonomous drones. So the one, the one is problem with ten thousand was you could, you could have, uh, so much place to launch there.

• 29:10Speaker 01

  Actually, it could, it takes, uh, much time to, um, prepare for launch. Two, three years, it's, it will be possible for now. I, I am really with launching, uh, five with one ground station drones, uh, at the one time.

• 29:31Speaker 01

  But I think it is possible to make, if it is fully autonomous drone, if you don't need to control drones,

• 29:40Speaker 01

  that it's a bit easier. As if you need to control any drones, for example, this 1100 drone, you need to control this in this time,

• 29:55Speaker 01

  this is much harder than make it fully autonomous.

• 30:00Speaker 00

  Yeah, so you say in two years it's possible that you would have, you would be able to launch 10,000 autonomous drones. I mean, you mean the tech will be there.

• 30:15Speaker 01

  Yes, so actually it could be even in this year, not about 10,000, but even if you could launch 100 drones.

• 30:27Speaker 01

  It's very hard to deal with it.

• 30:31Speaker 00

  I mean, do you worry though that Russia is going that direction?

• 30:38Speaker 01

  Everyone, everyone is going this direction. Russia, Ukraine, and anyone who wanted to have

• 30:49Speaker 01

  an up-to-date weapon, they are about fully autonomous weapons that can recognize the correct target. They are, that couldn't be jammed or something like that. Everyone wants this kind of weapon.

• 31:07Speaker 00

  Yeah, and the development, because, I mean, there is the hardware production, but a lot of that is software, is the AI.

• 31:16Speaker 01

  Yeah, of course, absolutely necessary.

• 31:20Speaker 00

  Yeah, and do you have AI developers that are working on models? Because that kind of model exists today.

• 31:31Speaker 01

  Yeah, yeah, yeah, yeah. We have not only one, many of them.

• 31:39Speaker 01

  The NATO countries are trying to get the right weapon, I mean, the right model for this. They are trying to learn, study this model correctly. USA, actually, anyone who wants this kind of drone are on the way to have this. They are trying to use AI, they are trying to

• 32:11Speaker 01

  place it on a drone and test it in our battleground.

• 32:18Speaker 00

  Yeah, I mean, the reason I ask is so much about this is the US has said that it will not develop fully autonomous lethal weapons for ethical reasons. But it seems on the battlefield,

• 32:40Speaker 00

  every side is trying to get an advantage, and that would be an obvious advantage. Do you, and you said everyone's working in that direction, and you said in two years you'll be able to field fully autonomous drones.

• 33:01Speaker 00

  I mean, that's a frightening world if two armies have huge numbers of fully autonomous drones that would kill every living thing in a certain zone. Is that where warfare is headed, do you think?

• 33:22Speaker 01

  Even in this century, we will see the conflict with, if it could be, if it enough conflict with

• 33:35Speaker 01

  some kind of big countries which have wanting to have very expensive weapons, it would.

• 33:46Speaker 01

  Yes, so this is just my opinion. But, you know, everyone knows it. Chinese are trying to get this kind of drones. They are testing their software here. They send people, their people here to get some kind of experience about truly the rules of the war just to

• 34:21Speaker 01

  change it.

• 34:26Speaker 01

  Just, just for example, the heavy armored vehicle like tanks. Russians, for example, in their assaults don't use the tanks as attacks. This is armored vehicle, just armored vehicle, nothing. They take some infantry on them, and the one reason is if you have a big, big, big jammer on it and move as far as is it go. That's all. Tank even isn't a tank. Actually, the artillery is suffering from drones. Infantry is suffering from drones. Everything is suffering from drones.

• 35:08Speaker 01

  And as a war we would have the rules. This war provides new rules

• 35:17Speaker 01

  of what we are doing, what we are afraid of. Is some kind of autonomous drone is present here. Of course, it's very good at recognizing this vehicle. But it is not good enough to know is it enemy or not. That's a problem. And this is what is much easier for Russia. If you don't care about is it enemy or just vehicle, you can just hit. And this is a point. They have drones with AI, just like you to we or something like that. But they don't have many of them.

• 36:01Speaker 01

  They have many stupid, stupid drones that you should have a radar, does it move, it recognize the target, hit the target. But

• 36:17Speaker 01

  they don't have many of smart drones. The Bumblebee is smart drone. But Russia don't have something like that.

• 36:27Speaker 00

  I would guess that that technology is going to be on the battlefield very soon if it's not there already. Is that, is that fair to say?

• 36:41Speaker 01

  Yeah, as I saw later,

• 36:45Speaker 01

  the drone with AI is on the battleground for now. They are present here. But they couldn't recognize with 100% accuracy that it is target where what you should hit. And this is a problem because

• 37:06Speaker 01

  they, they could use just civilian cars, just civilian car and use it as your own. So would you hit it or wouldn't you hit it? The drone just have no information. It is a source of car, they recognize this as a car. But the operator should understand is it enemy car? For example, if you are

• 37:32Speaker 01

  if you have a rear which where is no civilians, absolutely no civilians. You can use autonomous mission, any car, just hit it. But if you know that there is some civilians could be there or you couldn't use this kind of missions.

• 37:50Speaker 00

  Does Azov have just one R and D group or many R and D groups around the country?

• 37:58Speaker 01

  Actually, for now, many units have R and D groups with different, with different results, I would say. This way, with different results, some of them are good, some of them are not so good. They are trying to find their ways. For example, you know, though, as I know,

• 38:26Speaker 01

  I am on the call and some other R and D group is present.

• 38:36Speaker 01

  I know some people from other R and D groups. They are doing their stuff, we are doing our stuff. Sometimes we are doing

• 38:49Speaker 01

  in communication and as usually we are different departments.

• 38:57Speaker 00

  Yeah, and can I ask, in your group or is the primary focus navigation, jamming or object recognition? I mean, which is most important? I mean, target recognition.

• 39:18Speaker 01

  I understand. Actually, we are, for now, we are focused on navigation. We have some good results in it. I would also tell about this. We are focused on it, we are making some results. About some kind of recognition,

• 39:42Speaker 01

  this is an accept. When you can move to this area, then you can use some kind of recognition. For now, we are focused on navigation. If we are talking about autonomous mission.

• 39:59Speaker 01

  With them just, first of all, navigation then. Of course, this navigation is enough for reconnaissance drones. And this, it could do some missions even without recognition and then we just work with this thing.

• 40:21Speaker 00

  Yeah, and again, on the models, the AI models, you don't want to say where you're getting them? Is that right? And on the chips, what, where are you getting the semiconductors to run the models?

• 40:42Speaker 01

  Actually, this is not a secret. Everyone uses some kind of Chinese semiconductors. But now we are trying to move on Ukrainian and we actually have good results in it. We have actually even many kinds of absolutely Ukrainian drones with AI in it or without.

• 41:12Speaker 01

  And we have a good drones from our partners. Actually, it's absolutely true. They, actually, NATO countries are very, very good at software part. But as you understand, the problem of automations is mostly software. The NATO countries is much, much closer to have absolutely autonomous weapon as an AI.