Episode 25

Network Analysis for LoRaWAN - Dimitris Mamalis at Kudzu

Dimitris Mamalis, co-founder and CEO of Kudzu Technologies, talks about redefining how we understand and manage LoRaWAN networks. With roots in embedded systems and early experience using LoRa before the LoRaWAN protocol even existed, Dimitris brings a sharp, systems-level perspective to the challenges of deploying reliable long-range networks.

In this episode, he explains why most people deploying LoRaWAN aren’t just installing applications—they’re operating networks, whether they know it or not. Kudzu’s platform helps these accidental network operators understand what’s happening in the field, providing not just dashboards but actionable reports and monthly engineering-grade analysis to help fix what’s broken.

We dive into how Kudzu handles LoRaWAN network analysis, from initial simulation and digital modeling to in-field validation and optimization. Dimitris discusses how AI is used not to replace expert insight but to arm clients with better context, making problem-solving faster and more efficient. He also shares how the Helium boom was a proving ground for Kudzu’s capabilities, revealing common issues and accelerating the company’s growth.

Finally, Dimitris outlines how Kudzu continues to support Helium—not as a primary network, but as a roaming option. By helping customers identify coverage gaps and intelligently extend networks using Helium without additional gateways, Kudzu is building smarter, more resilient LoRaWAN deployments.

Dimitris on LinkedIn

Kudzu - LoRaWAN Analytics

  • Helium Global IoT Coverage - Want to know if Helium coverage exists where you need it? Check out this map!
  • Helium Foundation - The Helium Foundation's IoT Working Group (IOTWG) has generously provided support for the first 6 months of shows, please go check them out and consider using the Helium LoRaWAN as a primary or backup on your next deployment. With over a quarter million gateways deployed worldwide, it's likely that you have and can use Helium coverage.
  • Support The Show - If you'd like to support the MetSci Show financially, here's where you can donate on a one-time or an ongoing basis. Thank you!
  • MetSci Show - If you'd like to use our IoT or AI Data Value calculators, or you'd like to contact me, the MetSci Show site is the best way to do it.
  • MeteoScientific Console - Use LoRaWAN - The MeteoScientific Console allows you to use LoRaWAN today. As long as you have Helium coverage (and you probably do, about 90% of populated areas in the world have a gateway within 2 miles), you can onboard a sensor. You can always check coverage at https://explorer.helium.com and switch to the "IoT" tab in the top right.

Transcript
Speaker:

Today's guest on MeteoScientific's

The Business of LoRaWAN

2

:

is Dimitris Mamalis, co-founder

and CEO of Kudzu Technologies.

3

:

Dimitris

been working with LoRa since:

4

:

before the LoRaWAN protocol

was even standardized.

5

:

What started as a hardware

focused company has evolved into

6

:

one of the most advanced LoRaWAN

analytics platforms in the field.

7

:

Kudzu helps network operators,

whether they realize they’re operators to

8

:

design, monitor and troubleshoot networks

that power everything from water meters

9

:

to smart city infrastructure

In this conversation, we cover how Kudzu

10

:

tools give network

11

:

owners actionable insights, like whether

a gateway outage is critical or not?

12

:

And why most LoRaWAN deployments

break down not because of the technology

13

:

itself, but because of poor planning

or misconfigured devices.

14

:

Dimitris explains how LoRas connectionless

15

:

nature introduces

both power and complexity,

16

:

and why it's essential to treat

17

:

even small deployments like real networks

with real operational needs.

18

:

We also dig into how a Helium boom helped

Kudzu sharpen its ability

19

:

to explain network complexity

to non-experts, and what Dimitris sees

20

:

as the missing piece to take LoRaWAN

from hobbyist success to global

21

:

infrastructure standard: Third party

network quality certification.

22

:

This episode is sponsored by the Helium

Foundation's IoT Working Group.

23

:

Helium offers global LoRaWAN coverage

and is used by everyone

24

:

from hobbyists to businesses

deploying countrywide networks.

25

:

If you'd like to see if Helium

coverage exists near you,

26

:

check out the links in the show notes

to get started using Helium today.

27

:

You can sign up for a console account

28

:

with MeteoScientific

at console.meteoscientific.com.

29

:

Now let's dig into the

conversation with Dimitris.

30

:

Dimitris,

thanks so much for coming on the show.

31

:

Super excited to have you. It's

very nice being here.

32

:

It was a nice meeting you in Barcelona

conference and you know, setting this up.

33

:

We love to be here.

34

:

It's super cool.

35

:

I mean, we met during the Helium craze,

36

:

but I'd like to go back

a little bit further in.

37

:

Kudzu land and walk me through

what was the arc of development for Kudzu?

38

:

Did you guys start out thinking

you were going to build hardware

39

:

and then go into network analytics? Like,

what did that look like?

40

:

Well, the idea, the thing is

that we're both embedded system engineers.

41

:

So we did start from building our own

hardware and stuff like that.

42

:

It's still there.

43

:

But what has really pulled us into this

conversation and everything,

44

:

this technology

was actually the LoRa network business

45

:

in terms of trying to set up

nice networks, monitor networks

46

:

and stuff that we're going to talk,

I guess, later in this.

47

:

Yeah.

48

:

So you started with hardware

and it was it the Sprout that started

49

:

that still exists.

50

:

It's still exist.

51

:

It's just a little bit

52

:

less market ized, let's say,

because it's a kind of hard thing to do.

53

:

And also the idea and the business

behind the LoRaWAN analytics platform

54

:

has kind of skyrocketed for us last year,

especially after Helium.

55

:

So, yeah, we just decided to focus on this

a little bit more.

56

:

Cool. That makes sense.

57

:

I mean, gateway hardware is hard.

58

:

Okay, so if a customer comes to you,

59

:

what do they kind of see and do and get

when it comes to network analytics?

60

:

Like what should they expect?

61

:

Well, long story short,

let me just give you kind of explanation

62

:

of how we treat this thing.

63

:

We believe that we have this notion

that everyone who installs

64

:

more than 2 or 3 LoRaWAN gateways

for us is a network operator.

65

:

So when they come to us, they expect to

find services for designing these networks

66

:

and also monitoring

and evaluating the quality of the network.

67

:

After the installation

and after everything turns on.

68

:

That's the main idea behind it.

69

:

Okay.

70

:

And I would think with LoRaWAN,

just because it's such a robust network,

71

:

you can throw a couple gateways

up in a neighborhood and you're good.

72

:

But it sounds like from what

you guys have seen, there are maybe

73

:

significant improvements that you can make

in positioning those gateways

74

:

as well as some other factors. Is that

is that right?

75

:

Or is it that's kind of

are you guys overcomplicating anything?

76

:

This is our big, talk also within,

you know, internal in the LoRa Alliance.

77

:

We've been a LoRa Alliance member

for like 3 or 4 years now.

78

:

I had the chance to be there

in the beginning.

79

:

I was using LoRa back in 2012

for like three years before the wide

80

:

area network protocol was released,

and we could see coming.

81

:

We could see two important things,

two things that

82

:

I thought they were very important.

83

:

And this is how I started Kudzu.

84

:

One is that with that connectionless

technology,

85

:

literally a technology,

it's a little bit like walkie talkies.

86

:

So you talk, you're not connected.

87

:

Everybody can hear everything.

88

:

Of course

there's encryption and stuff like that,

89

:

and nobody else

can talk at the same point.

90

:

And we turn this into a wider networks

and installing like thousands,

91

:

hundreds of thousands of devices.

92

:

So I have the idea then,

you know, we've got a business there.

93

:

And the second thing was that

LoRa was promoted as a name,

94

:

an off the shelf technology.

95

:

Exactly as you describe.

96

:

You can grab the gateways, install them

somewhere, have very good coverage

97

:

and stuff like that.

98

:

But does this scale correct in

this is also a big question

99

:

that goes horizontally all around the IoT,

the ecosystem.

100

:

And so we as I said, we consider people

with gateways as network operators.

101

:

Although a lot of times in LoRa, people

consider themselves

102

:

application installers

so they don't install networks.

103

:

It's all water meters.

104

:

But to operate water meters

you need a network.

105

:

And we trying to give those people tools

to manage the networks

106

:

because they are not

and they don't need to be.

107

:

When it comes to LoRa telecom experts,

that's the idea.

108

:

So we try to engulf into the platform

all the expertise

109

:

you need in order to run networks

as they're supposed to be running.

110

:

And that's pretty cool

because I always thought of LoRaWAN is

111

:

as being this really high bar, like you do

need to be an expert in everything

112

:

to get to get anything.

113

:

That’s the story with embedded in IoT

a little bit, right?

114

:

I mean, it's that side

of, let's say full stack developer wants

115

:

you have things on the field

you need to do to know a little bit

116

:

about everything, right?

117

:

Yeah.

118

:

There's always a scenario

you don't speak C++,

119

:

you don't speak Python 7 or whatever

the latest crazy thing is like, oh,

120

:

for Christ's sake,

I just want this thing to work.

121

:

Okay,

so you guys do the network analysis piece.

122

:

Can you walk me through what

123

:

it might look like to do some fine

tuning of a LoRaWAN deployment?

124

:

So I've got five gateways up.

125

:

You come into my business and you're like,

here's what I think we can do.

126

:

Here's what you could see.

127

:

Well, the funny thing is

that you have five gateways

128

:

and then you come to our business

129

:

because you start seeing things going

not going as you expect.

130

:

Don't get me misunderstood.

131

:

It's not like LoRa has inherited problems,

132

:

but things scale

now of course becomes really large.

133

:

To be honest,

if you come with five gateways,

134

:

we'll probably just consult you for free

to give you the idea about how it is.

135

:

But if you come with let’s

say, a city level network or some kind of

136

:

infrastructure monitoring applications

and stuff like that, things can get messy.

137

:

One of the main things that we see

is, well, okay,

138

:

let me start from the beginning.

I started from the first step.

139

:

The first step,

of course, is designing the network.

140

:

Up to today, what we've seen,

these are people, again, with this idea

141

:

that this is an easy to install and easy

to deploy, technology.

142

:

They do RF analysis,

so they take their positions to the place

143

:

where they can pull their antennas

and see the propagation of the network

144

:

coverage and stuff like that.

145

:

That's super cool and super useful.

146

:

But this is just the first step.

It's completely different.

147

:

You design a network for general use

148

:

or a network for water meters

that are placed in Europe.

149

:

They're usually deep down, you know,

on the street level and stuff like that.

150

:

And it's completely different to design

a network for mobile applications,

151

:

for sensors moving around or for light

bulb control and stuff like that.

152

:

So we take all of these things

into consideration at the state level,

153

:

at the design level, the first state,

once you turn on your networks also.

154

:

Yeah.

155

:

Keep in mind that let me add here

that we don't give you a report

156

:

about the possible RF analysis.

157

:

We give you a digital model

that you can start verifying by the time

158

:

your network is, actually running.

159

:

So we give you this digital model

of our simulation.

160

:

And if you continue

to the rest of the services we have

161

:

we can start mapping the real scenario.

162

:

What you see on the field,

163

:

once your network

is there and compare it with a simulation,

164

:

it gives us a headstart

165

:

in the fixing issues

even before the whole network is deployed.

166

:

Because as you can imagine, a network,

167

:

especially in a city environment,

is not deploying within a day, right?

168

:

It takes some time.

169

:

So we can see these issues very early,

that the coverage

170

:

for this type of application

might not be as expected.

171

:

When we simulated the network, because

simulation, it's just a simulation.

172

:

Now this is the first idea.

173

:

Now moving on.

174

:

We do all this

kind of qualitative analysis.

175

:

I usually say that if you ask 20 people

what's a good LoRaWAN

176

:

network for you, they're going to give you

40 different answers.

177

:

There's no strict side line.

178

:

I've seen all the type of weird things

like one of our customers

179

:

a few months ago was asking for packet

error rate.

180

:

It's like it doesn't exist

181

:

in connectionless network,

so you can't really use this as a metric.

182

:

But they do think they want this.

183

:

So we do kind of an analysis

of your needs.

184

:

And we designed the KPIs.

We have some of our own KPIs.

185

:

And we also designed the KPIs

per application.

186

:

In order to give you an idea

about how your network works.

187

:

Of course these can be in dashboards,

engineers love dashboards.

188

:

Personally, I find it overwhelming.

189

:

A lot of times, you know, people

see graphs, but graphs don't

190

:

really mean anything in the end of the day

unless you're an expert.

191

:

So we do take the extra step,

and we do two things

192

:

which we think are pretty valuable.

193

:

You get an engineering report

at the end of the month, about

194

:

how good is your network

in terms of quality.

195

:

And you also get actionable items,

actionable alerts.

196

:

So we're going to tell you that

let me put it this way.

197

:

A typical, let's say, gateway management

platform will give you an idea,

198

:

will tell you that a gate was down.

199

:

What could do is it's gonna tell you

is that whether this is a critical event,

200

:

whether this is something you should draw

in on a Sunday afternoon, when go and fix

201

:

or whether this is something that you can,

you know, relax.

202

:

There's redundancy here and there.

203

:

So you are okay with it.

204

:

Last thing is that there's

a lot of debugging happening there.

205

:

As I told you in the beginning,

if you have like five or 10 or 50 yet,

206

:

which is very possible

that you will come to us

207

:

because you see some strange decision

in your network

208

:

and it's extremely difficult to debug.

209

:

It's just the sensors in the field.

210

:

Is it the gateways? Is it the server?

211

:

So we've asked you can really even,

you know, question the

212

:

AI that we've trained on this type of data

and you can really get answers

213

:

very quickly about what

might be going on in your network.

214

:

So that's the idea as funnily enough,

that was my next question.

215

:

Was this AI integration piece.

216

:

I think there's a a bigger conversation

about, you know, will

217

:

AI fix the overall IoT

kind of integration issue.

218

:

But in your case, is an AI just going

to put you out of a job where you

219

:

you have this AI and then I can talk to

and be like, hey, what should I do?

220

:

And it can know enough to say,

oh, well, you know, move over, Dimitris

221

:

this is what you need to do

to get around this.

222

:

And how close are you to that?

223

:

I'm glad that we've seen

it being used by our customers.

224

:

It's a little bit.

225

:

Well, I don't know

if it's gonna throw us out of business.

226

:

Be my guest. I'd love to be here in my

island doing something else,

227

:

and no problem

with that as long as they pay for it.

228

:

But it does give a lot of insights

for our customers, especially when they

229

:

have issues where they end up coming to us

for the consultation space,

230

:

because a lot of times

they're going to see issues,

231

:

but they need more input where our experts

on that, so they end up with us,

232

:

you know, doing meetings, weekly meetings

to to solve issues and stuff like that.

233

:

They come way more informed.

234

:

So they have an idea about

what's going on in the network already.

235

:

And we think that this is

this speeds up the process extremely

236

:

at the end of the day,

to get more value out of the product

237

:

because they can solve their issues

much faster.

238

:

Yeah, that makes sense.

239

:

So one of the things that I'm hearing

is this idea that you're explaining

240

:

what is a technical thing,

network analysis and understanding to

241

:

maybe non-technical experts

when it comes to network analysis.

242

:

And my guess is that your interaction with

Helium helped you guys get a lot better

243

:

at explaining non-technical or semi

technical things to non-technical people.

244

:

Can you tell me a little bit

about the experience with Helium back in?

245

:

What is it 2021

when the whole thing exploded?

246

:

It was a crazy idea. Yes.

247

:

Well, back in 2021,

248

:

the platform was already there,

but it was a less complete state.

249

:

So let's say.

250

:

But we already had integrated

this type of planning software

251

:

so you could actually do everything

on your browser.

252

:

We had all the 3D models in the browser

and stuff like that,

253

:

and you could directly just drag

and drop your antennas, your locations.

254

:

And I remember that Helium was coming

and I called my colleague and co-founder

255

:

Ioannis, and I was like, man,

we have a chance here.

256

:

And I didn't really I'm,

257

:

you know, I'm more of an engineer

and less of a salesman person.

258

:

My idea was that this is the time

to test our planning capabilities.

259

:

So we injected the Helium

algorithms about, how was it called?

260

:

I don't even remember. Sorry for this.

261

:

It was like, HIP 17.

262

:

All the different parts

about how you should are.

263

:

We kept integrating all this kind of stuff

and people were getting crazy.

264

:

And yes, as you very nicely put down,

we figured it out that

265

:

of course it was a commercial success

for the platform,

266

:

but it was also a huge success for us

because we sold exactly this.

267

:

We realized how this type of thing,

for what type of problems

268

:

people have with these networks

269

:

and how we can integrate

all these problems to the platform

270

:

and how we can explain this explosion

to the platform.

271

:

And it was a very, very,

very helpful experience for us.

272

:

By the way, we're still using

the Helium network here in the island.

273

:

So we have like a few hundreds of sensors

running.

274

:

So yeah, it's good to it's still going

I mean, it's surprising

275

:

to a lot of people to realize that Helium

not only is still going,

276

:

but it's it's still the largest

global LoRaWAN and still works.

277

:

That's true.

278

:

And we do have customers

which had this request.

279

:

We had foreseen this request,

but it came very late.

280

:

Some of our customers

281

:

have their own private networks,

not on Helium, but they do use Helium

282

:

where they needed to, where they have like

some kind of white holes in the network.

283

:

And they usually come to us

so that we can detect where they need to

284

:

and change their network

by not adding gateways, but instead buying

285

:

the data from Helium. And that's amazing.

286

:

I think it's a very

good and very nice solution.

287

:

And I think

288

:

that's part of Alper,

the CEO of the LoRa Alliance.

289

:

One of his big plans is to say, hey,

we should all be interchangeable.

290

:

So whether you're on Helium

or TTN or actually or whatever,

291

:

the thing is like, great.

292

:

They all talk to each other,

they all do all the kind of horse

293

:

trading behind the scenes as far as

who gets what for what data pack.

294

:

And but from the customer's

perspective of having coverage,

295

:

if there's any LoRaWAN coverage there,

it should be used and we're not there yet.

296

:

But I think that's exactly

297

:

that's also what we keep pushing down,

let's say the throat of the Greek market.

298

:

LoRa projects are being installed here

mainly due to the advancements

299

:

in water metering.

300

:

So what they're metering might belong to

different municipalities in the same area,

301

:

and we keep telling them

do not install more networks.

302

:

There is coverage around you can install

a couple of gate was here and there 10, 20

303

:

whatever.

304

:

But you can check the coverage of other

networks and start roaming the data.

305

:

This kind of this the dream of LoRa.

306

:

And they sometimes

307

:

also use the platform in order

to see if they have roaming capacity.

308

:

Let's say if they can allow other traffic

to go through their network

309

:

and stuff like that. And it's amazing.

310

:

I see it's pretty slow to pick up

because they haven't learned

311

:

how to, you know, collaborate yet,

but I hope it will be there soon.

312

:

Yeah. Yeah, that's that's my hope as well.

313

:

So what else do you see

314

:

coming down the pipe in LoRaWAN

that is relevant to what Kudzu is doing.

315

:

What are you excited about.

316

:

Oh, it's off the hook.

317

:

That's an unexpected question

to be honest.

318

:

We would love to see some kind

of certification on the network level.

319

:

An example

I usually give people that LoRa is on

320

:

right

now, the stage where you buy your sensors.

321

:

So it's like you buy your mobile phone,

your expensive certified mobile phone,

322

:

and you cannot really be sure

if the network is going to be there.

323

:

It's like the mobile telephone back

the 90s, you didn't have coverage

324

:

everywhere, didn't know what's going on.

325

:

We'd love to have some kind of,

well, certification, strict guideline

326

:

or whatever

in order to allow the network to be,

327

:

let's say criticized based on the quality

and ranked based on the quality.

328

:

And I think that would give a great level

of people to home, because to be honest,

329

:

LoRa is amazing.

330

:

But out in the market, you will see.

331

:

You will already know that, that

when people have some problems,

332

:

they first blame the technology

but not their installations.

333

:

And this is what we want to change

in order to, you know, leverage

334

:

the whole thing.

335

:

I have been so many times

that I blamed LoRa,

336

:

and it turns out that I put in the app

key wrong or the DevEUI or something.

337

:

So you can imagine what happens

when you do this.

338

:

And they support like a hundred thousand

sensors, right?

339

:

It's like, yeah, things can collapse.

340

:

Okay.

341

:

So maybe one thing we

we might see in the future and I'm totally

342

:

putting words in your mouth here.

343

:

Is that Kudzu or something

344

:

like it comes out with the global

LoRaWAN network coverage tool.

345

:

And so you can look and see

hey what's what's covered in my area.

346

:

Is it going to be

Helium is going to be TTN, or Actility.

347

:

That would be amazing to be honest.

348

:

Yeah, that would be a really nice thing

to do, to have at least for everyone.

349

:

Okay.

350

:

Well,

the next time we have you back on there,

351

:

you should have it ready, right? Yep.

352

:

There's a task for the next time.

353

:

Yeah, yeah, I don't

I give you about six months.

354

:

Dimitris.

thanks so much for making the time.

355

:

Come on. I know you're super busy

running a business.

356

:

I appreciate you, carving out some time

and talking to me about network analysis

357

:

when it comes to LoRa.

Thank you so much, and I hope.

358

:

I guess we'll see you on, smart

cities down November, right?

359

:

That's right.

360

:

Yeah. See you there.

361

:

That's it for this episode

of The Business of LoRaWAN.

362

:

I built this for you.

363

:

So whether you're a business owner,

a LoRaWAN professional, or a hobbyist,

364

:

the intent is to give you great LoRaWAN

information.

365

:

Of course,

the best information doesn't come from me.

366

:

It comes from the conversations

367

:

we have with the people building

and deploying this tech in the real world.

368

:

And that's where you come in.

369

:

LoRaWAN is a global

patchwork of talent and ideas.

370

:

And ironically,

for a globally connected network,

371

:

most of the brilliant folks

working on it are connected yet.

372

:

Help me change that.

373

:

Introduce me

to someone awesome in your network.

374

:

Someone doing meaningful work in Laurette.

375

:

Or just shoot me a name.

376

:

I'll take it from there

377

:

and get them on the show

so we can share their work with the world.

378

:

You can always find me at metsci.show

379

:

That's M-E-T-S-C-I dot

380

:

S-H-O-W, metsci.show.

381

:

If you want to support the show

in other ways, you can subscribe.

382

:

Leave a review,

share it with your corner of the world.

383

:

All those are super helpful.

384

:

If you'd like to support financially,

you can go to support.metsci.show

385

:

for both one time and recurring options.

386

:

We're also open to sponsors.

387

:

If your company serves

the LoRaWAN community

388

:

and you want to reach this dedicated

audience, let's talk.

389

:

If you want to try out LoRaWAN for

yourself, create a MeteoScientific account

390

:

at console.meteoscientific.com

and get your first 400 DC for free,

391

:

which is enough to run a device

sending hourly for about a year.

392

:

This show is supported

393

:

by a grant from the Helium Foundation

and produced by Gristle King, Inc..

394

:

I'm Nik Hawks.

I'll see you in the next show.

All Episodes Previous Episode

About the Podcast

Show artwork for The Business of LoRaWAN
The Business of LoRaWAN
Learn From the Pros

Listen for free

About your host

Profile picture for Nik Hawks

Nik Hawks

Incurably curious, to stormy nights and the wine-dark sea!