WEBVTT
Language: en-US

1
00:00:04.280 --> 00:00:09.680
<v Speaker 1>This is obviously one way where you try to
solve a problem that is not solvable with

2
00:00:09.680 --> 00:00:10.880
<v Speaker 1>classical automation.

3
00:00:11.159 --> 00:00:15.819
<v Speaker 1>The nice thing is that you usually have a
pretty defined business case already.

4
00:00:16.559 --> 00:00:20.450
<v Speaker 1>The downside is that it's usually a pretty
complex problem, otherwise they would

5
00:00:20.450 --> 00:00:21.909
<v Speaker 1>have solved it already.

6
00:00:22.690 --> 00:00:23.129
<v Speaker 1>Um.

7
00:00:24.309 --> 00:00:28.180
<v Speaker 1>Then the other way around, if you go top
down, um, like.

8
00:00:29.229 --> 00:00:32.770
<v Speaker 2>Welcome to the trend detection podcast
powered by Siemens.

9
00:00:32.950 --> 00:00:34.909
<v Speaker 2>And now your host Nel Sullivan.

10
00:00:37.069 --> 00:00:41.470
<v Niall Sullivan>Hello and welcome to the latest episode of
the Trend detection podcast.

11
00:00:41.869 --> 00:00:45.470
<v Niall Sullivan>Uh, I'm really pleased to be joined by a
brand new guest, cos I often have a lot of

12
00:00:45.470 --> 00:00:50.060
<v Niall Sullivan>recurring guests on this podcast, but a
brand new guest and one of my, Um, Siemens

13
00:00:50.060 --> 00:00:53.939
<v Niall Sullivan>colleagues, Christian Zillner is joining
us today, and we're talking about the very

14
00:00:53.939 --> 00:00:58.279
<v Niall Sullivan>important topic of industrial AI, but
we're actually, I know it's one of these

15
00:00:58.279 --> 00:01:02.060
<v Niall Sullivan>phrases which is often used, but we're
gonna sort of ground it in some reality,

16
00:01:02.259 --> 00:01:05.900
<v Niall Sullivan>particularly on how to deploy it and get
real value from it ultimately is what we're

17
00:01:05.900 --> 00:01:06.339
<v Niall Sullivan>looking at.

18
00:01:06.459 --> 00:01:09.459
<v Niall Sullivan>Um, but before we sort of dive into that,
I'll hand over to Christian.

19
00:01:09.809 --> 00:01:11.830
<v Niall Sullivan>Just to introduce yourself and welcome to
the podcast.

20
00:01:12.569 --> 00:01:12.879
<v Speaker 1>Hi.

21
00:01:13.000 --> 00:01:13.690
<v Speaker 1>Hi, everyone.

22
00:01:14.010 --> 00:01:15.250
<v Speaker 1>Um, my name is Chris.

23
00:01:15.370 --> 00:01:16.029
<v Speaker 1>Thanks for having me.

24
00:01:16.290 --> 00:01:18.199
<v Speaker 1>As I already mentioned, I work for
Siemens.

25
00:01:18.269 --> 00:01:22.699
<v Speaker 1>I'm heading the global AI deployment team
in DI automation.

26
00:01:23.169 --> 00:01:28.529
<v Speaker 1>Means that I am in touch with a lot of
projects and use cases where we deploy AI on

27
00:01:28.529 --> 00:01:33.889
<v Speaker 1>the shop floor at our customer premises
together with partners in this fast moving

28
00:01:33.889 --> 00:01:34.230
<v Speaker 1>environment.

29
00:01:34.529 --> 00:01:37.889
<v Speaker 1>It's a lot of fun, a lot of learnings, and
I'm very happy to be here.

30
00:01:38.180 --> 00:01:42.260
<v Niall Sullivan>I guess just a question sort of a little
bit off topic for what we um sort of

31
00:01:42.260 --> 00:01:44.629
<v Niall Sullivan>planned, but you mentioned all the
different news cases.

32
00:01:44.699 --> 00:01:47.959
<v Niall Sullivan>Is that a good thing to have lots of
different news cases to focus on, or is it, you

33
00:01:47.959 --> 00:01:51.019
<v Niall Sullivan>know, can it be quite challenging from
that perspective, cos usually people say

34
00:01:51.019 --> 00:01:52.819
<v Niall Sullivan>focus on things, right?

35
00:01:53.319 --> 00:01:55.739
<v Speaker 1>Yeah, focus definitely is always a good
idea.

36
00:01:56.160 --> 00:02:00.319
<v Speaker 1>I think from my perspective we're just
working in so many different verticals with

37
00:02:00.319 --> 00:02:01.790
<v Speaker 1>different requirements.

38
00:02:02.419 --> 00:02:07.879
<v Speaker 1>Some are very close to our shop floor
where we're having what we call adaptive

39
00:02:07.879 --> 00:02:08.279
<v Speaker 1>control.

40
00:02:08.460 --> 00:02:11.800
<v Speaker 1>This means like AI really controls
processes on the shop floor.

41
00:02:11.899 --> 00:02:16.750
<v Speaker 1>They are very different requirements to, I
don't know, some monitoring or predictive

42
00:02:16.750 --> 00:02:20.100
<v Speaker 1>maintenance of pumps in a, I don't know,
pipeline.

43
00:02:20.949 --> 00:02:24.089
<v Speaker 1>Of gas or whatever, uh, you want to
monitor.

44
00:02:24.289 --> 00:02:29.529
<v Speaker 1>Uh, so this is the variety we have to
handle on, on the siemen side, but it's also

45
00:02:30.029 --> 00:02:33.350
<v Speaker 1>very interesting to always see the same
learnings all over.

46
00:02:33.690 --> 00:02:33.699
<v Speaker 1>Yeah.

47
00:02:34.259 --> 00:02:35.610
<v Speaker 1>Our focus is always a good idea.

48
00:02:35.990 --> 00:02:36.949
<v Niall Sullivan>Yeah, exactly.

49
00:02:37.110 --> 00:02:39.880
<v Niall Sullivan>Maybe I already knew the answer to that
question, I guess, but I thought it was

50
00:02:39.880 --> 00:02:39.910
<v Niall Sullivan>interesting.

51
00:02:40.125 --> 00:02:43.794
<v Niall Sullivan>Interesting to ask anyway, but, um, and,
and I guess today we're actually looking at

52
00:02:43.794 --> 00:02:48.154
<v Niall Sullivan>that journey from an AI lab, um, all the
way through to the shop floor, and I guess

53
00:02:48.154 --> 00:02:50.054
<v Niall Sullivan>the value sort of generated there.

54
00:02:50.315 --> 00:02:54.595
<v Niall Sullivan>And we're also focusing, I mentioned in
the intro, um, around deployment as well.

55
00:02:54.714 --> 00:02:58.514
<v Niall Sullivan>So what, what does that actually mean to
you, Christian, in practice?

56
00:02:58.755 --> 00:02:59.005
<v Niall Sullivan>Um,

57
00:02:59.690 --> 00:03:05.570
<v Speaker 1>Yeah, it's a very big word obviously, but
for us what we realized and learned over

58
00:03:05.570 --> 00:03:10.770
<v Speaker 1>time is that deployment really is working
with our customers because usually an

59
00:03:10.770 --> 00:03:16.210
<v Speaker 1>implementation of AI or scaling of AI on
the shop floor where you need to bridge IT

60
00:03:16.210 --> 00:03:19.369
<v Speaker 1>and OT interests is kind of a change
process.

61
00:03:19.759 --> 00:03:26.360
<v Speaker 1>We usually trigger a lot of responsibility
processes, scoping of business cases,

62
00:03:26.699 --> 00:03:28.830
<v Speaker 1>financial decisions attached to it.

63
00:03:29.110 --> 00:03:33.669
<v Speaker 1>And then obviously there's also the
technical implementation, but it's just a part

64
00:03:33.669 --> 00:03:36.190
<v Speaker 1>of this whole deployment cycle we're in.

65
00:03:36.619 --> 00:03:43.500
<v Speaker 1>And obviously we always try to to do that
in line with partners and internal teams

66
00:03:43.500 --> 00:03:44.580
<v Speaker 1>and then customer teams.

67
00:03:44.660 --> 00:03:49.820
<v Speaker 1>So it's quite a bit of a tricky, complex
environment we're in there, but it's also

68
00:03:49.820 --> 00:03:50.979
<v Speaker 1>very interesting, obviously.

69
00:03:51.179 --> 00:03:53.979
<v Niall Sullivan>I noticed you've mentioned sort of
partners a couple of times already.

70
00:03:54.100 --> 00:03:56.809
<v Niall Sullivan>So what's a partner's role in these
projects?

71
00:03:56.820 --> 00:03:59.880
<v Niall Sullivan>What both their importance and, you know,
what, what do they bring to the table?

72
00:04:00.220 --> 00:04:06.216
<v Speaker 1>What we learned over time is that it is
crucial if you want To really scale and I

73
00:04:06.216 --> 00:04:09.345
<v Speaker 1>think you called it from the lab to the
shop floor.

74
00:04:09.585 --> 00:04:15.095
<v Speaker 1>If you really want to implement real
industrial AI use cases, you need to have an

75
00:04:15.095 --> 00:04:17.036
<v Speaker 1>owner of the application.

76
00:04:17.216 --> 00:04:23.406
<v Speaker 1>What we mean by owner is that someone who
services over the lifetime of the

77
00:04:23.406 --> 00:04:27.635
<v Speaker 1>application, it's no different to any
automation project you're probably used to,

78
00:04:27.835 --> 00:04:31.135
<v Speaker 1>and those system integrators, um, machine
builders.

79
00:04:31.782 --> 00:04:38.492
<v Speaker 1>Global solution integrators are our
partners on that journey to implement, and this

80
00:04:38.492 --> 00:04:39.842
<v Speaker 1>is the one kind of partner.

81
00:04:39.851 --> 00:04:42.131
<v Speaker 1>And then obviously we also have technology
partners.

82
00:04:42.171 --> 00:04:47.611
<v Speaker 1>We partner with camera vendors, we partner
with software, other software players.

83
00:04:47.652 --> 00:04:51.721
<v Speaker 1>We even partner with cloud providers where
we try to standardize the deployment

84
00:04:51.721 --> 00:04:52.541
<v Speaker 1>architecture.

85
00:04:52.812 --> 00:04:57.652
<v Speaker 1>So there's ecosystem is really, really
important to make it a successful

86
00:04:57.652 --> 00:04:58.451
<v Speaker 1>implementation.

87
00:04:59.320 --> 00:05:03.040
<v Niall Sullivan>I mean, you mentioned successful
implementations, but I mean, quite often, and

88
00:05:03.040 --> 00:05:06.720
<v Niall Sullivan>you'll be able to, you know, tell me and
the listeners, but, better than anyone, you

89
00:05:06.720 --> 00:05:11.929
<v Niall Sullivan>know, that a lot of industrial AI
copilots, sorry, um, show a lot of promise, but,

90
00:05:12.040 --> 00:05:14.429
<v Niall Sullivan>you know, don't scale at the end of it,
unfortunately.

91
00:05:14.519 --> 00:05:18.260
<v Niall Sullivan>I mean, so from what you've seen, you
know, what, what are the key reasons behind

92
00:05:18.260 --> 00:05:18.589
<v Niall Sullivan>that?

93
00:05:18.799 --> 00:05:21.920
<v Niall Sullivan>You know, what, and maybe also on the flip
side, what, you know, where do you see

94
00:05:21.920 --> 00:05:24.589
<v Niall Sullivan>the success factors, what success factors
are involved?

95
00:05:24.839 --> 00:05:28.239
<v Speaker 1>Oh, yeah, um, we can confirm, like, um, as
Siemens.

96
00:05:28.359 --> 00:05:31.160
<v Speaker 1>A lot of own factories ourselves.

97
00:05:31.200 --> 00:05:36.609
<v Speaker 1>I was responsible for implementing AI in
our own factories as well, and we failed

98
00:05:36.609 --> 00:05:37.440
<v Speaker 1>many times.

99
00:05:38.510 --> 00:05:43.829
<v Speaker 1>My very subjective learnings there are
that expectation management is key in

100
00:05:43.829 --> 00:05:49.040
<v Speaker 1>combination with the business case because
usually those use cases develop over

101
00:05:49.040 --> 00:05:49.359
<v Speaker 1>time.

102
00:05:49.519 --> 00:05:55.369
<v Speaker 1>So you need to have a good backing and
quite a bit of a business case to finance an

103
00:05:55.369 --> 00:05:57.440
<v Speaker 1>industrial AI use case.

104
00:05:57.959 --> 00:06:04.049
<v Speaker 1>And a second, we bet in the, uh, in the
past, a lot of custom

105
00:06:04.049 --> 00:06:09.320
<v Speaker 1>architectures and we heavily regretted it
afterwards in the scaling phase because if

106
00:06:09.320 --> 00:06:15.519
<v Speaker 1>you're locked into custom solutions, it's
really hard to enable operation services

107
00:06:15.519 --> 00:06:21.299
<v Speaker 1>that are mentioned before for partners
taking over the solution, um, doing MLOps

108
00:06:21.299 --> 00:06:26.160
<v Speaker 1>services, um, providing it, uh, is really,
really, really hard if you don't have any

109
00:06:26.160 --> 00:06:26.709
<v Speaker 1>standards.

110
00:06:26.910 --> 00:06:30.309
<v Niall Sullivan>And so what I'm hearing there is when
these projects don't get off the ground, is it

111
00:06:30.309 --> 00:06:32.579
<v Niall Sullivan>usually a technology problem or, or
something else?

112
00:06:32.589 --> 00:06:35.269
<v Niall Sullivan>You've sort of mentioned, I think, you
know, with the architecture you mentioned

113
00:06:35.269 --> 00:06:37.459
<v Niall Sullivan>there, but are there other factors at play
as well?

114
00:06:37.670 --> 00:06:37.750
<v Niall Sullivan>Yeah,

115
00:06:37.829 --> 00:06:44.690
<v Speaker 1>Obviously the evolution of model
availability in our space is just enormous, so

116
00:06:44.920 --> 00:06:50.825
<v Speaker 1>we A lot more possibilities in the
accessibility towards machine learning platforms

117
00:06:50.825 --> 00:06:54.554
<v Speaker 1>to train models has evolved a lot in the
last couple of years.

118
00:06:54.744 --> 00:07:01.214
<v Speaker 1>So most of the predictive quality
inspection vision use cases also robotic physical

119
00:07:01.214 --> 00:07:03.404
<v Speaker 1>eye use cases we see and we implement.

120
00:07:03.940 --> 00:07:06.350
<v Speaker 1>The model side is pretty much covered.

121
00:07:06.619 --> 00:07:11.950
<v Speaker 1>The second bigger hurdle is obviously then
data availability, and this connected to

122
00:07:11.950 --> 00:07:17.709
<v Speaker 1>a standardized approach of how to collect
data, how to provide it for model

123
00:07:17.709 --> 00:07:19.829
<v Speaker 1>training, and then obviously connect them.

124
00:07:19.934 --> 00:07:26.015
<v Speaker 1>Models on the shop floor with the real
sensors is a challenge, because there's no

125
00:07:26.015 --> 00:07:31.855
<v Speaker 1>standards for IoT platforms yet that
standardize this whole process from cloud to

126
00:07:31.855 --> 00:07:36.015
<v Speaker 1>edge across all verticals and all use
cases we see that.

127
00:07:36.329 --> 00:07:38.820
<v Speaker 1>So this is definitely quite a bit of a
challenge.

128
00:07:39.160 --> 00:07:43.880
<v Speaker 1>And then the third thing is the deployment
platform, uh, where we, where we have,

129
00:07:44.040 --> 00:07:47.540
<v Speaker 1>where we see great progress with GPUs
availability.

130
00:07:47.920 --> 00:07:52.149
<v Speaker 1>But there, this is connected then a lot
with, uh, with the price point and the

131
00:07:52.149 --> 00:07:56.679
<v Speaker 1>business case I mentioned before already,
because you can spend quite a bit of on

132
00:07:56.679 --> 00:08:00.140
<v Speaker 1>compute if you want, either in the cloud
or on the edge.

133
00:08:00.829 --> 00:08:04.179
<v Niall Sullivan>I mean, maybe we're leaning more into sort
of the adoption of technology here, but

134
00:08:04.179 --> 00:08:07.709
<v Niall Sullivan>are there cultural factors that complement
what you, you were explaining there on

135
00:08:07.709 --> 00:08:08.690
<v Niall Sullivan>the technology side?

136
00:08:09.549 --> 00:08:15.350
<v Speaker 1>I think you have to have a look on who
drives the business case or who drives the

137
00:08:15.350 --> 00:08:16.070
<v Speaker 1>use case.

138
00:08:16.510 --> 00:08:22.630
<v Speaker 1>Obviously the profile and the capabilities
of an OT team or shop floor team is very

139
00:08:22.630 --> 00:08:26.269
<v Speaker 1>different to maybe a data science IT
minded team.

140
00:08:26.739 --> 00:08:32.780
<v Speaker 1>And for an industrial AI project or a
deployment or a scale, even a scaling, you

141
00:08:32.780 --> 00:08:38.169
<v Speaker 1>need different profiles, and we always
differentiate them on first, the domain

142
00:08:38.169 --> 00:08:40.479
<v Speaker 1>expert that really knows the process well.

143
00:08:40.580 --> 00:08:46.299
<v Speaker 1>He's usually responsible for the business
case, seeing the implications, safety,

144
00:08:46.419 --> 00:08:46.580
<v Speaker 1>quality.

145
00:08:46.885 --> 00:08:47.744
<v Speaker 1>All of this.

146
00:08:48.025 --> 00:08:50.025
<v Speaker 1>Then we have the IT personas in it.

147
00:08:50.625 --> 00:08:55.744
<v Speaker 1>It's a lot about security, CRA compliance,
cyber resilient Act compliance is a big,

148
00:08:55.945 --> 00:08:57.424
<v Speaker 1>big issue these days.

149
00:08:57.784 --> 00:09:01.864
<v Speaker 1>Then we have the data science team doing
the data engineering, um, selecting the

150
00:09:01.864 --> 00:09:07.174
<v Speaker 1>model, model validation, and then you need
some kind of connectivity guy that

151
00:09:07.174 --> 00:09:11.984
<v Speaker 1>bridges the gap from like data collection
on the shop floor, bringing, bringing it

152
00:09:11.984 --> 00:09:15.255
<v Speaker 1>to the cloud or to the computer
environment where you train the model, and then.

153
00:09:15.729 --> 00:09:20.330
<v Speaker 1>Reverse it back and bounce the model back
to the shop floor then and the fifth

154
00:09:20.330 --> 00:09:24.969
<v Speaker 1>obviously is once you've delivered the
project is then service where you need guys

155
00:09:24.969 --> 00:09:27.330
<v Speaker 1>to take care of the envelope services.

156
00:09:27.530 --> 00:09:32.690
<v Speaker 1>So it's quite a bit of a cultural aspect
in it because you have different personas.

157
00:09:33.010 --> 00:09:39.419
<v Speaker 1>Um, and we try to, to reduce the
complexity by a lot with UX or UI.

158
00:09:39.609 --> 00:09:45.590
<v Speaker 1>So G AI also helps us a lot to reduce
complexity in the interaction of engineering,

159
00:09:45.890 --> 00:09:49.650
<v Speaker 1>but we're still not there in the whole
life cycle or MLlos life cycle.

160
00:09:49.750 --> 00:09:49.849
<v Speaker 1>Yeah,

161
00:09:50.010 --> 00:09:51.450
<v Niall Sullivan>maybe let's dive insight a little bit.

162
00:09:51.489 --> 00:09:55.750
<v Niall Sullivan>You said generative AI is helping to make
the complex less complex.

163
00:09:55.869 --> 00:09:57.210
<v Niall Sullivan>in, in, in what ways?

164
00:09:57.280 --> 00:09:59.070
<v Niall Sullivan>What, what examples can you provide?

165
00:10:00.289 --> 00:10:04.659
<v Speaker 1>Usually an OT engineer doesn't have any
Python programming skills.

166
00:10:04.929 --> 00:10:11.650
<v Speaker 1>So if we have a copilot kind of interface
where he can do intent-based

167
00:10:11.650 --> 00:10:16.719
<v Speaker 1>engineering for, I don't know, some kind
of a, just making up an example, a vision

168
00:10:16.719 --> 00:10:21.109
<v Speaker 1>solution where he wants to detect
scratches on a metal sheet or whatever.

169
00:10:21.500 --> 00:10:27.239
<v Speaker 1>Then intent-based engineering could be a
key to enable an automation guy to set up

170
00:10:27.239 --> 00:10:33.059
<v Speaker 1>and, and sketch and develop an industrial
AI solution and train a model without any

171
00:10:33.059 --> 00:10:35.200
<v Speaker 1>data science, uh, know-how.

172
00:10:35.739 --> 00:10:41.789
<v Speaker 1>Uh, so this enables us to, to scale and
the, the cool thing is, If we could do that,

173
00:10:41.880 --> 00:10:46.549
<v Speaker 1>we could reintegrate in the already
existing service processes because automation

174
00:10:46.549 --> 00:10:47.559
<v Speaker 1>knowledge is there.

175
00:10:47.880 --> 00:10:54.440
<v Speaker 1>So talking again about machine learning
operations, if some guy or if an industrial

176
00:10:54.440 --> 00:11:01.400
<v Speaker 1>AI solution only requires automation
knowledge, we are a lot further down the road

177
00:11:01.400 --> 00:11:02.440
<v Speaker 1>in AI adoption.

178
00:11:03.010 --> 00:11:06.890
<v Niall Sullivan>And I know we're sort of jumping back to
what you were saying before, those, those

179
00:11:06.890 --> 00:11:11.289
<v Niall Sullivan>different personas, is it, it's
interesting because with so many different, you

180
00:11:11.289 --> 00:11:14.330
<v Niall Sullivan>know, personas involved in a, in a
deployment.

181
00:11:14.409 --> 00:11:18.530
<v Niall Sullivan>Is there someone who actually fully owns
the deployment from your perspective, or is

182
00:11:18.530 --> 00:11:20.909
<v Niall Sullivan>it a shared ownership model?

183
00:11:21.210 --> 00:11:22.919
<v Niall Sullivan>What, how would you sort of describe it?

184
00:11:23.169 --> 00:11:28.609
<v Speaker 1>It obviously always depends a lot on the
use case, but we always say that IT owns

185
00:11:28.609 --> 00:11:30.429
<v Speaker 1>the standards and the security.

186
00:11:30.780 --> 00:11:37.140
<v Speaker 1>And OT owns the use case because they, the
intensivation is quite different.

187
00:11:37.419 --> 00:11:43.559
<v Speaker 1>IT is Incentivized on security and OT is
incentivized on

188
00:11:43.559 --> 00:11:47.929
<v Speaker 1>efficiency, output, up time, all of those.

189
00:11:48.440 --> 00:11:54.599
<v Speaker 1>So this is how we usually split it and
then if we translate what you just asked is

190
00:11:54.599 --> 00:12:00.460
<v Speaker 1>that the use case usually stays with the
OT, which means that the solution has to be

191
00:12:00.840 --> 00:12:04.400
<v Speaker 1>serviced by the OT department or business
unit.

192
00:12:04.869 --> 00:12:08.510
<v Niall Sullivan>And is that ownership always clear from
the start within an organization?

193
00:12:08.549 --> 00:12:12.750
<v Niall Sullivan>I know it depends, of course, between
organizations and what use case, but, you

194
00:12:12.750 --> 00:12:16.630
<v Niall Sullivan>know, is it always clear straight away,
this, these are the defined roles, or can it

195
00:12:16.630 --> 00:12:18.169
<v Niall Sullivan>get a bit blurred sometimes?

196
00:12:19.190 --> 00:12:20.369
<v Speaker 1>Ah, it's very blurry.

197
00:12:20.669 --> 00:12:26.309
<v Speaker 1>Usually we have to kind of, we have those
C-level things where two C-level guys say,

198
00:12:26.830 --> 00:12:29.570
<v Speaker 1>oh, we want to do something with AI on the
shop floor.

199
00:12:30.200 --> 00:12:36.530
<v Speaker 1>Um, or we have those bottom-up, uh, use
cases where a guy tried everything from

200
00:12:36.530 --> 00:12:41.590
<v Speaker 1>classical automation to smart sensors,
smart cameras, whatever, and then ends up at

201
00:12:41.590 --> 00:12:46.369
<v Speaker 1>AI vision, for example, because he could
not find another way to solve his problem,

202
00:12:46.489 --> 00:12:51.530
<v Speaker 1>but it's a significant problem, so he
wants to, to throw some money at it or it's

203
00:12:51.530 --> 00:12:52.969
<v Speaker 1>significant business wise for him.

204
00:12:53.359 --> 00:12:58.909
<v Speaker 1>So those are two very different directions
and, and either of them has different

205
00:12:59.369 --> 00:13:02.849
<v Speaker 1>responsibilities, budget constraints, and
all of it.

206
00:13:03.330 --> 00:13:08.809
<v Speaker 1>But as part of our deployments, we usually
develop racing matrix uh with our, with

207
00:13:08.809 --> 00:13:14.090
<v Speaker 1>our customers to see the responsibilities,
accountabilities within the shop floor

208
00:13:14.090 --> 00:13:16.929
<v Speaker 1>deployment and then obviously services is
one of them.

209
00:13:17.739 --> 00:13:22.020
<v Niall Sullivan>Um, and I, but I guess to sort of shift
gears a little bit, you know, another thing

210
00:13:22.020 --> 00:13:25.679
<v Niall Sullivan>I wanted to cover today is sort of the
difference, if we're looking at industrial AI

211
00:13:25.679 --> 00:13:30.299
<v Niall Sullivan>in terms of the deployment of these
projects, you know, cloud versus edge, and where

212
00:13:30.299 --> 00:13:34.080
<v Niall Sullivan>you see things at the minute is, you know,
is it leaning towards one or the other,

213
00:13:34.140 --> 00:13:38.419
<v Niall Sullivan>or, you know, are organizations thinking,
yeah, in, in a certain way around this

214
00:13:38.419 --> 00:13:38.619
<v Niall Sullivan>topic?

215
00:13:38.710 --> 00:13:39.919
<v Niall Sullivan>Maybe you could go into that a little bit.

216
00:13:40.630 --> 00:13:46.200
<v Speaker 1>Yeah, I think AI, the adoption is heavily
structured in three maturity processes.

217
00:13:46.359 --> 00:13:48.119
<v Speaker 1>One is model availability.

218
00:13:48.280 --> 00:13:53.320
<v Speaker 1>There we did already quite a huge progress
in the last couple of years.

219
00:13:53.549 --> 00:13:55.159
<v Speaker 1>Then we have data availability.

220
00:13:55.520 --> 00:13:57.590
<v Speaker 1>This is heavily connected to Aure cloud.

221
00:13:57.880 --> 00:14:01.260
<v Speaker 1>And then we have the deployment or the
inference.

222
00:14:01.559 --> 00:14:06.200
<v Speaker 1>Maybe you heard of that buzzword a lot
like where to deploy my AI on.

223
00:14:06.450 --> 00:14:10.250
<v Speaker 1>And data and inference is a different
thing.

224
00:14:10.450 --> 00:14:16.409
<v Speaker 1>So I think you need quite a bit of a
hybrid approach to collect data and then push

225
00:14:16.409 --> 00:14:22.030
<v Speaker 1>it to the engineering environment where
you have the possibility to train your

226
00:14:22.030 --> 00:14:23.750
<v Speaker 1>models, validate all of it.

227
00:14:24.169 --> 00:14:29.549
<v Speaker 1>This is one thing, and then you need to
choose a deployment environment.

228
00:14:29.849 --> 00:14:33.630
<v Speaker 1>So where you want to do the influencing of
your AI model.

229
00:14:34.140 --> 00:14:38.030
<v Speaker 1>And this is very dependent on your, on
your use case.

230
00:14:38.210 --> 00:14:45.150
<v Speaker 1>If you have very high latency and very
high um security uh requirements, uh, you

231
00:14:45.150 --> 00:14:46.809
<v Speaker 1>usually go edge only.

232
00:14:47.390 --> 00:14:52.309
<v Speaker 1>If you have less latency requirements,
like if you do monitoring of a pipeline, just

233
00:14:52.309 --> 00:14:57.710
<v Speaker 1>going back to my, to my example, where you
maybe do one or you have to do one

234
00:14:57.710 --> 00:15:02.030
<v Speaker 1>prediction a minute, then you can easily
also go cloud.

235
00:15:02.710 --> 00:15:05.229
<v Speaker 1>A second factor is the data format.

236
00:15:05.429 --> 00:15:10.909
<v Speaker 1>If you want to push all your data to the
cloud, the cloud providers will love it

237
00:15:10.909 --> 00:15:13.260
<v Speaker 1>because it would have been enormously
expensive.

238
00:15:13.270 --> 00:15:14.950
<v Speaker 1>If you talk about vision data.

239
00:15:15.270 --> 00:15:19.349
<v Speaker 1>We have use cases where you have 16
cameras to connect it to.

240
00:15:19.429 --> 00:15:25.330
<v Speaker 1>One AI model, so it will be large amount
of, of data streams, um, that has to be

241
00:15:25.330 --> 00:15:31.270
<v Speaker 1>orchestrated and they usually we, we go to
Edge and then just have like data

242
00:15:31.270 --> 00:15:35.330
<v Speaker 1>collectors that, that collect or sort the
data that you want to push to the cloud

243
00:15:35.330 --> 00:15:35.969
<v Speaker 1>and later on.

244
00:15:36.789 --> 00:15:41.710
<v Niall Sullivan>So that, so that's the, the way around,
you know, sort of rising, the rising costs

245
00:15:41.710 --> 00:15:44.969
<v Niall Sullivan>of cloud or that could potentially be from
that particular use case you mentioned.

246
00:15:45.849 --> 00:15:46.309
<v Niall Sullivan>Yeah.

247
00:15:46.890 --> 00:15:49.690
<v Niall Sullivan>Take a step back a bit because I know
we've talked about use cases and clearly

248
00:15:49.690 --> 00:15:53.460
<v Niall Sullivan>that's the important part, because before
you deploy AI you need to know sort of

249
00:15:53.460 --> 00:15:56.010
<v Niall Sullivan>what use case you or problem you're
looking to solve.

250
00:15:56.130 --> 00:15:58.409
<v Niall Sullivan>So, what does that look like in your, in
your world?

251
00:15:58.559 --> 00:16:02.510
<v Niall Sullivan>Sort of, what's the process for defining
use cases with a, with a customer?

252
00:16:03.210 --> 00:16:04.929
<v Speaker 1>I think it's 22 ways.

253
00:16:05.169 --> 00:16:10.409
<v Speaker 1>Um, the one way is, uh, if you have like
a, a shop floor problem that you want to

254
00:16:10.409 --> 00:16:15.710
<v Speaker 1>solve because it's immensely important for
The factory output or

255
00:16:17.289 --> 00:16:19.890
<v Speaker 1>and you go and basically bottom up.

256
00:16:20.289 --> 00:16:25.640
<v Speaker 1>This is obviously one way where you try to
solve a problem that is not solvable with

257
00:16:25.640 --> 00:16:26.849
<v Speaker 1>classical automation.

258
00:16:27.130 --> 00:16:31.789
<v Speaker 1>The nice thing is that you usually have a
pretty defined business case already.

259
00:16:32.479 --> 00:16:36.400
<v Speaker 1>The downside is that it's usually a pretty
complex problem, otherwise they would

260
00:16:36.400 --> 00:16:38.080
<v Speaker 1>have solved it already.

261
00:16:38.940 --> 00:16:39.140
<v Speaker 1>Um.

262
00:16:40.260 --> 00:16:45.580
<v Speaker 1>Then the other way around, if you go top
down, um, like headquarter down, it's

263
00:16:45.580 --> 00:16:48.729
<v Speaker 1>obviously the, the scaling is a lot
different.

264
00:16:48.859 --> 00:16:54.820
<v Speaker 1>Um, if we, if we, for example, take, take
a vertical like consumer packaged goods,

265
00:16:54.859 --> 00:16:56.890
<v Speaker 1>you have a lot of standardization.

266
00:16:57.140 --> 00:17:03.580
<v Speaker 1>We usually look at very standardized
processes and try to get into those standard

267
00:17:03.580 --> 00:17:10.285
<v Speaker 1>architectures, or line setups, the, um,
the AI or the industrial AI solution or Use

268
00:17:10.285 --> 00:17:15.344
<v Speaker 1>case already implemented and standardized
because then you have a copy and paste

269
00:17:15.344 --> 00:17:16.625
<v Speaker 1>approach while scaling.

270
00:17:16.944 --> 00:17:22.785
<v Speaker 1>Keep in mind here to choose the
architecture upfront, because then if you want to

271
00:17:22.785 --> 00:17:27.864
<v Speaker 1>scale and it's custom and not serviceable,
it's obviously the cost of failure is a

272
00:17:27.864 --> 00:17:28.344
<v Speaker 1>lot higher.

273
00:17:28.505 --> 00:17:32.344
<v Niall Sullivan>Yeah, I guess you, you touched upon it
then there a little bit, Christian, but maybe

274
00:17:32.344 --> 00:17:33.545
<v Niall Sullivan>we could dive into that a bit more.

275
00:17:33.625 --> 00:17:37.375
<v Niall Sullivan>So we've talked about the, you know, the
problems of getting these projects sort of

276
00:17:37.375 --> 00:17:41.949
<v Niall Sullivan>off the ground, you know, maybe in a pilot
phase, but Equally, at scale, you know,

277
00:17:42.109 --> 00:17:46.869
<v Niall Sullivan>clearly success has been achieved on one
level, but what sort of issues become

278
00:17:46.869 --> 00:17:49.739
<v Niall Sullivan>clear, and you mentioned architecture
could be one, but what other sort of

279
00:17:49.739 --> 00:17:53.410
<v Niall Sullivan>challenges sort of appear as, um, these
projects scale upwards?

280
00:17:54.050 --> 00:17:59.739
<v Speaker 1>I think the first thing is always use case
definition, um, connected with the

281
00:17:59.739 --> 00:18:03.739
<v Speaker 1>business case, but this is no different to
any other automation project.

282
00:18:04.099 --> 00:18:05.859
<v Speaker 1>So don't be too scared here.

283
00:18:05.939 --> 00:18:12.819
<v Speaker 1>I think it's, it's the same, just do your
job and scoping the project and defining

284
00:18:12.819 --> 00:18:15.410
<v Speaker 1>success metrics, uh, scope of work, all of
that.

285
00:18:15.579 --> 00:18:19.359
<v Speaker 1>I think this is equally important for an
AI project or even more important.

286
00:18:19.859 --> 00:18:21.859
<v Speaker 1>Second, obviously is data availability.

287
00:18:22.219 --> 00:18:24.260
<v Speaker 1>Um, there's shit in, shit out.

288
00:18:24.380 --> 00:18:30.920
<v Speaker 1>If you have unreliable, uh, data that is
not standardized, it will be costly or even

289
00:18:31.540 --> 00:18:32.959
<v Speaker 1>not reliable in the end.

290
00:18:33.540 --> 00:18:35.540
<v Speaker 1>Third is model development.

291
00:18:35.660 --> 00:18:39.180
<v Speaker 1>Obviously, you need a good data science
partner or you do it yourself.

292
00:18:39.369 --> 00:18:44.819
<v Speaker 1>It, it's your choice as a, as a solution
owner if you want to own it and see the

293
00:18:44.819 --> 00:18:49.160
<v Speaker 1>model development as your core competency
or you want to have it in-house, uh, or

294
00:18:49.160 --> 00:18:50.300
<v Speaker 1>your, your partner here.

295
00:18:50.670 --> 00:18:54.260
<v Speaker 1>And then 4th is the deployment environment
as, as mentioned before.

296
00:18:54.390 --> 00:18:59.949
<v Speaker 1>Um, here, I think it's important, the one
side is the data connectivity to

297
00:18:59.949 --> 00:19:02.010
<v Speaker 1>standardize and productize this.

298
00:19:02.469 --> 00:19:07.750
<v Speaker 1>And I would strongly recommend you to
partner there, um, because connectivity is a

299
00:19:07.750 --> 00:19:09.479
<v Speaker 1>key issue for shopflow data.

300
00:19:09.910 --> 00:19:12.930
<v Speaker 1>And as it is critical, it's very
important.

301
00:19:12.989 --> 00:19:18.189
<v Speaker 1>And then the other bit is the inference
environment you want to choose, whether it's

302
00:19:18.189 --> 00:19:23.949
<v Speaker 1>in the cloud or on the shop floor with
Edge devices locally hosted or then in the in

303
00:19:23.949 --> 00:19:24.510
<v Speaker 1>the cloud.

304
00:19:25.390 --> 00:19:29.089
<v Niall Sullivan>And it, it's interesting that you, you
were sort of balancing, you know, whether to

305
00:19:29.089 --> 00:19:33.890
<v Niall Sullivan>own stuff for the organization to own,
own, own it or outsource that potentially.

306
00:19:34.109 --> 00:19:35.510
<v Niall Sullivan>And particularly on the, the model side.

307
00:19:35.550 --> 00:19:39.589
<v Niall Sullivan>I mean, do you have a recommendation
either way, or is it very much down to the, the

308
00:19:39.589 --> 00:19:42.469
<v Niall Sullivan>use case, the organization, how they want
to approach things?

309
00:19:42.550 --> 00:19:46.250
<v Niall Sullivan>There's no right or wrong, it's just
what's the right approach for them, maybe.

310
00:19:46.790 --> 00:19:50.000
<v Speaker 1>I think the right approach is what you can
service.

311
00:19:50.400 --> 00:19:55.089
<v Speaker 1>If you have global availability of data
science services within your corporation.

312
00:19:55.680 --> 00:19:58.479
<v Speaker 1>You can do it in-house if not partner.

313
00:19:59.020 --> 00:20:01.189
<v Niall Sullivan>Yeah, that, that keeps it very simple,
right?

314
00:20:01.290 --> 00:20:05.140
<v Niall Sullivan>When, you know, the, I guess the danger is
with some of this, well, a lot of this is

315
00:20:05.140 --> 00:20:07.420
<v Niall Sullivan>to overthink it, but that keeps it very
simple, so that's good.

316
00:20:07.699 --> 00:20:08.479
<v Niall Sullivan>That's very good.

317
00:20:08.699 --> 00:20:12.479
<v Niall Sullivan>And I guess going back to the, the scale
question as well, because you've mentioned

318
00:20:12.479 --> 00:20:16.339
<v Niall Sullivan>the importance, uh, of, you know,
consistency, particularly with architecture, and

319
00:20:16.339 --> 00:20:19.300
<v Niall Sullivan>I'm sure there's other, um, you know,
other areas too.

320
00:20:19.380 --> 00:20:21.859
<v Niall Sullivan>But I guess there's also got to be
elements of flexibility as well.

321
00:20:21.979 --> 00:20:26.140
<v Niall Sullivan>So I'm just interested in how you sort of
balance those two very contrasting things.

322
00:20:26.510 --> 00:20:28.890
<v Niall Sullivan>In a, in a project when scaling in
particular.

323
00:20:29.880 --> 00:20:31.369
<v Speaker 1>I think this is strongly connected.

324
00:20:31.489 --> 00:20:34.250
<v Speaker 1>In our case, we have the approach of
partnering.

325
00:20:34.810 --> 00:20:39.290
<v Speaker 1>This has always been our business also in
automation, where we have system

326
00:20:39.290 --> 00:20:43.930
<v Speaker 1>integrators, line builders as partners,
and it's the same for AI solutions.

327
00:20:44.010 --> 00:20:50.959
<v Speaker 1>We try to have a operator of like 80%
product and 20% solution, and this makes

328
00:20:50.959 --> 00:20:55.250
<v Speaker 1>it also an attractive business for our
partners and we are developing.

329
00:20:55.739 --> 00:20:59.900
<v Speaker 1>What we call verticalized use case
offerings there together with our partners where

330
00:20:59.900 --> 00:21:02.660
<v Speaker 1>we try to productize as much as possible.

331
00:21:03.050 --> 00:21:07.339
<v Speaker 1>Um, not saying that Siemens delivers all
of it, like, not all the hardware, for

332
00:21:07.339 --> 00:21:13.209
<v Speaker 1>example, if we go into AI vision, we, we,
we don't have, or, or we don't have any

333
00:21:13.209 --> 00:21:17.420
<v Speaker 1>camera hardware, for example, we partner
there as well, but we try to productize as

334
00:21:17.420 --> 00:21:20.619
<v Speaker 1>much as we can, between 70 and 90%.

335
00:21:20.910 --> 00:21:27.430
<v Speaker 1>And then we have always an adaptability
service or like a customized approach to, to

336
00:21:27.430 --> 00:21:30.329
<v Speaker 1>make it a great fit to the customer
requirements.

337
00:21:30.550 --> 00:21:30.709
<v Speaker 1>Yeah,

338
00:21:30.829 --> 00:21:31.300
<v Niall Sullivan>that makes sense.

339
00:21:31.349 --> 00:21:35.010
<v Niall Sullivan>That, that seems like a very good sort of
balance between, between the two.

340
00:21:35.430 --> 00:21:38.150
<v Niall Sullivan>You know, I could relate it back to sort
of Sensai, which we haven't talked about

341
00:21:38.150 --> 00:21:42.109
<v Niall Sullivan>predictive maintenance, and it's not
necessarily, um, we need to, but it just

342
00:21:42.109 --> 00:21:45.670
<v Niall Sullivan>reminds me of, you know, Sensai's great
value of being able to integrate with

343
00:21:45.670 --> 00:21:49.750
<v Niall Sullivan>existing technology on the shop floor and
working with partners and being open like

344
00:21:49.750 --> 00:21:51.530
<v Niall Sullivan>that, which again is, you know, seems
accelerator.

345
00:21:51.589 --> 00:21:52.949
<v Niall Sullivan>You could even go down that route as well.

346
00:21:52.989 --> 00:21:54.609
<v Niall Sullivan>It's all very much linked together.

347
00:21:55.680 --> 00:21:59.599
<v Niall Sullivan>So, as we, um, sort of come to the end,
we're gonna have a quick fire round in a

348
00:21:59.599 --> 00:22:03.439
<v Niall Sullivan>second, but before we do, I just wanted to
ask sort of one final question.

349
00:22:03.479 --> 00:22:08.280
<v Niall Sullivan>It's more about just whether you could
provide one sort of piece of advice for an

350
00:22:08.280 --> 00:22:11.719
<v Niall Sullivan>organization that's looking to move from
AI experimentation, because I'm sure it can

351
00:22:11.719 --> 00:22:16.040
<v Niall Sullivan>be tempting to stay in that phase, you
know, indefinitely, you know, to find the

352
00:22:16.040 --> 00:22:20.319
<v Niall Sullivan>perfect solution, um, to go from there,
all the way through to the large scale

353
00:22:20.319 --> 00:22:21.569
<v Niall Sullivan>deployments we've been talking about.

354
00:22:21.680 --> 00:22:24.900
<v Niall Sullivan>So what's that one piece of advice just to
sort of try and narrow it down?

355
00:22:26.189 --> 00:22:29.380
<v Speaker 1>Standardize your deployment platform right
from the start.

356
00:22:29.920 --> 00:22:36.400
<v Speaker 1>Cost of switching is super high, and I
would choose like max 3 scalable use cases

357
00:22:36.400 --> 00:22:42.040
<v Speaker 1>that are quite standardized in your, in
your shop floor environment and where, where

358
00:22:42.040 --> 00:22:47.400
<v Speaker 1>you can sync up on with the machine
learning operations partner that could help you

359
00:22:47.400 --> 00:22:48.099
<v Speaker 1>service it.

360
00:22:48.400 --> 00:22:50.760
<v Speaker 1>It can be an internal partner or an
external partner.

361
00:22:50.839 --> 00:22:53.160
<v Speaker 1>Internal partner, for example, could be
your IT team.

362
00:22:53.619 --> 00:22:58.329
<v Speaker 1>Um, that could take over or it's an
external partner that you work with.

363
00:22:58.650 --> 00:23:00.310
<v Niall Sullivan>Fantastic, fantastic, very clear.

364
00:23:00.469 --> 00:23:04.589
<v Niall Sullivan>So, the quickfire round, the, the name
says it all, it's a very quick answers to

365
00:23:04.589 --> 00:23:05.349
<v Niall Sullivan>quick questions.

366
00:23:05.469 --> 00:23:07.189
<v Niall Sullivan>Don't think about it too much.

367
00:23:07.829 --> 00:23:08.869
<v Niall Sullivan>So we've got previews.

368
00:23:08.949 --> 00:23:09.390
<v Niall Sullivan>Let's start.

369
00:23:09.489 --> 00:23:12.949
<v Niall Sullivan>So the biggest myth about deploying
industrial AI.

370
00:23:13.630 --> 00:23:20.069
<v Speaker 1>I think there's no one solution cover,
covers it all approach

371
00:23:20.349 --> 00:23:20.910
<v Speaker 1>possible.

372
00:23:21.709 --> 00:23:23.079
<v Niall Sullivan>Perfect, perfect.

373
00:23:23.459 --> 00:23:26.959
<v Niall Sullivan>And one thing every successful AI
deployment has in common,

374
00:23:27.430 --> 00:23:28.579
<v Speaker 1>a good business case.

375
00:23:29.410 --> 00:23:34.849
<v Niall Sullivan>And if you had, if you could choose
between cloud, edge, or a hybrid approach, if

376
00:23:34.849 --> 00:23:38.650
<v Niall Sullivan>you had to pick one, which one wins most
for Industrial AIA today?

377
00:23:39.790 --> 00:23:40.400
<v Speaker 1>Hybrid.

378
00:23:40.699 --> 00:23:41.569
<v Niall Sullivan>Best of both worlds,

379
00:23:41.619 --> 00:23:41.819
<v Speaker 1>of

380
00:23:41.819 --> 00:23:42.400
<v Niall Sullivan>course.

381
00:23:43.359 --> 00:23:43.550
<v Speaker 1>Best of both worlds.

382
00:23:44.420 --> 00:23:47.489
<v Speaker 1>Most expensive one, but maybe,

383
00:23:47.660 --> 00:23:51.619
<v Niall Sullivan>maybe, but no, fantastic, fantastic, we've
just introduced that quickfire round, so

384
00:23:51.619 --> 00:23:55.619
<v Niall Sullivan>I'm, I'm still sort of testing it out, but
it's, um, it's interesting what responses

385
00:23:55.619 --> 00:23:55.959
<v Niall Sullivan>we get.

386
00:23:56.060 --> 00:23:59.859
<v Niall Sullivan>But, um, but no, for today, I think we've,
we've come towards the end of our

387
00:23:59.859 --> 00:24:03.380
<v Niall Sullivan>conversation, Christian, it's been
fantastic to, to have you on, talk about this.

388
00:24:03.930 --> 00:24:06.869
<v Niall Sullivan>Very interesting and very broad subjects
as well.

389
00:24:07.119 --> 00:24:11.089
<v Niall Sullivan>Um, but I guess if people want to find out
a bit more about your work, is there

390
00:24:11.089 --> 00:24:12.650
<v Niall Sullivan>anywhere we can point people

391
00:24:12.650 --> 00:24:13.109
<v Speaker 1>to?

392
00:24:13.479 --> 00:24:16.589
<v Speaker 1>Yeah, I think the easiest is add me on
LinkedIn.

393
00:24:17.209 --> 00:24:23.670
<v Speaker 1>I think that's one, and, um, you can find
us on the Siemens website.

394
00:24:23.819 --> 00:24:28.890
<v Speaker 1>So obviously, there's a product called
Industrial AI Suite, uh, or if you want to

395
00:24:28.890 --> 00:24:31.989
<v Speaker 1>check out our AI vision offerings, they're
called Inspector.

396
00:24:32.310 --> 00:24:38.930
<v Speaker 1>Um, feel free to check them, and there you
can see how we are trying to simplify

397
00:24:39.030 --> 00:24:44.530
<v Speaker 1>AI or industrial AI to make it more
accessible for, for automation engineers.

398
00:24:44.910 --> 00:24:45.390
<v Niall Sullivan>Fantastic.

399
00:24:45.510 --> 00:24:49.430
<v Niall Sullivan>Yeah, so we'll put all those links in the,
in the, in the show notes afterwards in

400
00:24:49.430 --> 00:24:49.770
<v Niall Sullivan>that case.

401
00:24:49.989 --> 00:24:50.430
<v Niall Sullivan>Fantastic.

402
00:24:50.530 --> 00:24:52.250
<v Niall Sullivan>But um, We it for today.

403
00:24:52.510 --> 00:24:56.300
<v Niall Sullivan>Thank you so much for joining us,
Christian, on your first, uh, appearance on the,

404
00:24:56.310 --> 00:24:57.989
<v Niall Sullivan>um, trend detection podcast.

405
00:24:58.069 --> 00:25:00.890
<v Niall Sullivan>Hopefully not the last, and hopefully we
haven't scared you off too much.

406
00:25:02.739 --> 00:25:03.369
<v Speaker 1>Thanks a lot for

407
00:25:03.369 --> 00:25:03.849
<v Niall Sullivan>having me.

408
00:25:04.069 --> 00:25:05.540
<v Niall Sullivan>No, no, of course, anytime.

409
00:25:06.150 --> 00:25:09.900
<v Niall Sullivan>Thank you everyone for joining us, um, for
this episode, wherever you are, wherever

410
00:25:09.900 --> 00:25:12.260
<v Niall Sullivan>you're listening, and we'll see you on the
next episode.

411
00:25:12.630 --> 00:25:13.060
<v Niall Sullivan>Thank you.

412
00:25:15.020 --> 00:25:19.979
<v Speaker 2>Please like and subscribe on all major
podcast channels and find out more about

413
00:25:19.979 --> 00:25:23.359
<v Speaker 2>Sensei predictive maintenance at
Siemens.com.
