UX for AI

EP. 94 - The Startup That Turned Waste Into Resource w/ Umutcan Duman

• Bonanza Studios

Share episode

Send us a text

 Umutcan Duman, CEO of Evreka, joins us to share how a group of engineers turned waste collection into a global smart tech platform. From IoT sensors to AI-powered insights, they’ve been solving real problems long before the buzzwords. This is the story of building Evreka — one bin at a time. 

You can find Umutcan here: https://nl.linkedin.com/in/umutcanduman

Interested in joining the podcast? DM Behrad on LinkedIn:
https://www.linkedin.com/in/behradmirafshar/

This podcast is made by Bonanza Studios, Germany’s Premier Digital Design Studio:
https://www.bonanza-studios.com/

1
00:00:00,000 --> 00:00:10,040
Welcome to UX for AI. Omojan, thanks a lot for coming on the podcast. You're a busy man.

2
00:00:10,040 --> 00:00:16,680
I can see with the wealth and extent that you are behind Evrica. And yeah, we'd really

3
00:00:16,680 --> 00:00:25,360
love to use every minute of it and go through what you have created. It's a work of wonder

4
00:00:25,360 --> 00:00:32,920
if I would say it properly, maybe like give audience the way that you shared with me in

5
00:00:32,920 --> 00:00:38,020
the initial call. That was very beautifully put. I think let's start from there because

6
00:00:38,020 --> 00:00:43,080
I wanted to be on the recording and then we can take it further. I know that you are going

7
00:00:43,080 --> 00:00:47,080
to share some screens and also show us really cool stuff as well.

8
00:00:47,080 --> 00:00:52,760
Yeah, of course. So this is our tent here as Evrica. We started as four co-founders.

9
00:00:52,760 --> 00:00:57,800
Back then I was a student in my last year in university. So we were knowing each other

10
00:00:57,800 --> 00:01:04,200
from the university time, school friends. So the initial problem that we have discovered

11
00:01:04,200 --> 00:01:11,580
is the inefficient based collection operation. We are all engineers, Berkay, Matt, Mehmet,

12
00:01:11,580 --> 00:01:18,680
and myself. Unfortunately, we lost Berkay two years later than we started. He passed away.

13
00:01:18,680 --> 00:01:23,840
So we are also carrying his dreams with us, trying to reach the level that we would like

14
00:01:23,840 --> 00:01:31,400
to reach altogether from the beginning. Yeah, so the first day was with Berkay. So we actually

15
00:01:31,400 --> 00:01:37,560
waited for the waste collection truck to ask our questions to the truck driver because

16
00:01:37,560 --> 00:01:42,080
that was our assumption and that they were just visiting every waste bin every single

17
00:01:42,080 --> 00:01:47,840
day in a static structure. So it was actually proven that they are doing the same route

18
00:01:47,840 --> 00:01:53,920
every single day without knowing which bin is full, which bin is empty. So they do the

19
00:01:53,920 --> 00:01:59,040
same thing every day, which is actually a huge inefficiency. So that was the initial

20
00:01:59,040 --> 00:02:05,280
starting point for us because this was something that we have been learning in engineering,

21
00:02:05,280 --> 00:02:10,900
industrial engineering department, like route optimization, efficiency, all these things.

22
00:02:10,900 --> 00:02:15,760
So that was trying to implement what we have learned on the field. And waste management

23
00:02:15,760 --> 00:02:22,720
is actually a kind of old school industry. So basically, things have been in a certain,

24
00:02:22,720 --> 00:02:28,760
you know, structure for a long time. Yeah. So if you try to innovate, if you try to change

25
00:02:28,760 --> 00:02:33,280
something, this is a little difficult. But if you can manage it, then you can have a

26
00:02:33,280 --> 00:02:40,080
big impact on the outcome of the industry. So that was a big challenge. So this is how

27
00:02:40,080 --> 00:02:44,800
we have started and we get it together within like two and a half months. Matt and Mehmet

28
00:02:44,800 --> 00:02:51,080
also joined our team. As four co-founders, we have started. I was even trying to catch

29
00:02:51,080 --> 00:02:56,000
my exams because I was, you know, my friends were calling me. I was running to the exam

30
00:02:56,000 --> 00:03:03,680
from the office. So luckily, I, you know, I finished my study without losing years. And

31
00:03:03,680 --> 00:03:08,900
then yeah, here we are. I didn't know one of your co-founders passed away hits. Yeah.

32
00:03:08,900 --> 00:03:15,360
Sadly, it's a miss, but also it must fill your furnace, not to take time for granted

33
00:03:15,360 --> 00:03:20,960
and try to create a legacy that he would be proud of. Yeah. Hopefully, hopefully we can

34
00:03:20,960 --> 00:03:29,040
make him proud. So you know, waste management, and I studied a bit as well, seems like to

35
00:03:29,040 --> 00:03:36,240
be a straightforward topic. You need to reduce waste, but it's a very difficult engineering

36
00:03:36,240 --> 00:03:43,800
problem. For example, for one, if a truck would go around the route and there are like 400

37
00:03:43,800 --> 00:03:50,240
bins, how would you use technology to know which bins is full, which bins is half empty,

38
00:03:50,240 --> 00:03:56,140
which bin is empty? That's a big engineering problem in my opinion, as an engineer with

39
00:03:56,140 --> 00:04:02,240
the engineering background. So how did you go? Like it's a big complex ecosystem. Where

40
00:04:02,240 --> 00:04:08,260
did you start? And maybe like walk us through this past year, 10 year journey from a bird's

41
00:04:08,260 --> 00:04:14,560
eye viewpoint and break down the journey, the evolution of Africa. Yeah, of course.

42
00:04:14,560 --> 00:04:20,480
Well, when we have started, we were not we, but in general, the ecosystem were talking

43
00:04:20,480 --> 00:04:27,440
about garbage. Yeah. And then it became waste and then it became material and then it became

44
00:04:27,440 --> 00:04:34,240
resource. So it's basically kind of, you know, evolution of the industry as well. So it was

45
00:04:34,240 --> 00:04:40,240
not, it was not that much valuable industry back then, it was all like the logistics,

46
00:04:40,240 --> 00:04:45,840
just carrying the garbage from one point to another just to dump it or burn it or whatever.

47
00:04:45,840 --> 00:04:51,220
Of course, in some countries, you know, more mature approaches were there, but not, it

48
00:04:51,220 --> 00:04:58,020
was not like globally best practice. All countries or all cities or all the companies were following

49
00:04:58,020 --> 00:05:06,080
the best practices. So basically as a normal random citizen, we didn't know about commercial

50
00:05:06,080 --> 00:05:11,680
waste. We didn't know about industrial waste, hazardous waste. Those are different verticals

51
00:05:11,680 --> 00:05:17,040
of waste management. As a normal citizen, we were only aware of the municipal waste

52
00:05:17,040 --> 00:05:25,040
management. So we only thought the municipality is our customer. Right. Well, actually this

53
00:05:25,040 --> 00:05:29,920
was not the case most of the time, because although municipalities are the responsible,

54
00:05:29,920 --> 00:05:35,840
generally these operations are managed by the private companies on behalf of cities.

55
00:05:35,840 --> 00:05:41,720
It's like outsourced operations. Right. So we were trying to optimize the routes of waste

56
00:05:41,720 --> 00:05:48,240
collection trucks for municipal waste operations. That was the initial starting point. Then,

57
00:05:48,240 --> 00:05:54,440
obviously, we realized there are different verticals of waste management and it is not

58
00:05:54,440 --> 00:05:59,160
only collection operations. You, you need to also manage the fleet. You need to also

59
00:05:59,160 --> 00:06:04,800
manage the employees. You also need to manage the different relationships, structures like

60
00:06:04,800 --> 00:06:11,340
invoicing proof of services, all of these. And it's not about collection only, managing

61
00:06:11,340 --> 00:06:19,080
the material after collection is another half of the waste management vertical. Yeah. So

62
00:06:19,080 --> 00:06:25,960
collection and material recovery. They call it material recovery facility in general.

63
00:06:25,960 --> 00:06:30,840
You can do different kinds of activities related with the material. You can just wash it, you

64
00:06:30,840 --> 00:06:37,280
can bail it, you can crash it. You can do many different types of activities after collection

65
00:06:37,280 --> 00:06:43,260
of the material. You got even harder problem to solve. Yeah, we are actually not satisfied

66
00:06:43,260 --> 00:06:49,560
with the problem that we solve. Then we always try to make it a bigger problem and then solve

67
00:06:49,560 --> 00:06:55,720
it. I think this is kind of the trigger, inner trigger for us, as we try to find it a bigger

68
00:06:55,720 --> 00:07:00,940
problem, more difficult problem to solve it. So, yeah, related to route optimization, it's

69
00:07:00,940 --> 00:07:07,040
actually a big problem because route optimization is one type of problem and route optimization

70
00:07:07,040 --> 00:07:12,760
within waste management is actually a different and a little more complex problem because

71
00:07:12,760 --> 00:07:18,600
you have different time windows. So those trucks cannot travel like 24 hours. That's right,

72
00:07:18,600 --> 00:07:24,520
that's right. Some of those trucks cannot get into some of the streets. So you need

73
00:07:24,520 --> 00:07:31,100
to use different vehicles for different types of streets and some different containers should

74
00:07:31,100 --> 00:07:36,760
be managed by different types of vehicles because some of the containers are big. Some

75
00:07:36,760 --> 00:07:43,700
of the containers are small. Too many variables. Yeah, it's getting more complex. So many variables.

76
00:07:43,700 --> 00:07:49,260
Variables can be easy, but constraints are getting also complex. So you have many different

77
00:07:49,260 --> 00:07:53,860
variables, so you have many different constraints and trying to manage them every day. And if

78
00:07:53,860 --> 00:07:58,960
you have different shifts, also this is bringing another complexity. Sometimes it also makes

79
00:07:58,960 --> 00:08:03,940
it simpler. But yeah, that was a big challenge. So we said it is a big enough challenge to

80
00:08:03,940 --> 00:08:10,540
tackle. Yeah, for engineers, massive complex ecosystem, why not? Yeah, this is what we

81
00:08:10,540 --> 00:08:17,600
enjoy. So this is a recipe for AI, especially from the traditional sense, machine learning,

82
00:08:17,600 --> 00:08:24,080
then now AI, especially gen AI. So how, so you started way back, then there was any gen

83
00:08:24,080 --> 00:08:31,200
AI or large language models were widespread used. How did you leverage, especially machine

84
00:08:31,200 --> 00:08:38,500
learning before the whole LLMs and gen AI came to being? Yeah. So I think this is like

85
00:08:38,500 --> 00:08:44,420
different. We have a different approach. Why? Because we studied in Ankara, it's the capital

86
00:08:44,420 --> 00:08:49,160
of Turkey. And our university is Middle East Technical University. This is like one of

87
00:08:49,160 --> 00:08:55,140
the best engineering university in Ankara. And there are different kinds of like two

88
00:08:55,140 --> 00:09:00,760
different kinds of approach in Turkey. One, like Ankara approach, more like engineering,

89
00:09:00,760 --> 00:09:07,460
trying to solve the problem and Istanbul approach is more on the commercial side, you know,

90
00:09:07,460 --> 00:09:13,380
so we solve the big problems, but we are not good at selling and marketing, packaging it.

91
00:09:13,380 --> 00:09:18,440
Istanbul is actually really good at selling and packaging and marketing. So we actually

92
00:09:18,440 --> 00:09:25,440
didn't know about IoT when we were starting. We thought if we can understand the fullness

93
00:09:25,440 --> 00:09:31,060
level of the bin, then we can do route optimization for waste collection trucks. Then we said,

94
00:09:31,060 --> 00:09:36,200
okay, let's try to find a solution to understand how full is the bin. It's not about, you know,

95
00:09:36,200 --> 00:09:40,940
trying to become an IoT company. It's about trying to find a solution for a problem that

96
00:09:40,940 --> 00:09:48,680
we see. So you were IoT before IoT? Before IoT was a thing. Well, we were not aware of

97
00:09:48,680 --> 00:09:54,820
IoT when we had built the field level sensor. Then IoT became a thing. Then we said, hey,

98
00:09:54,820 --> 00:09:59,520
we are actually an IoT company. But we didn't know that we are becoming an IoT company when

99
00:09:59,520 --> 00:10:04,620
we are building this sensor. That's fantastic. This field level sensors is patented, but

100
00:10:04,620 --> 00:10:11,080
is your patent or is something that that? It's not something that you can get a patent on,

101
00:10:11,080 --> 00:10:19,140
but you can have this design ownership kind of thing because it's... Got it. Then we realized

102
00:10:19,140 --> 00:10:26,040
the field level sensor topic is going to become kind of just a standard device. Right. So

103
00:10:26,040 --> 00:10:32,280
we need to provide a software solution which can work with different kinds of sensors too.

104
00:10:32,280 --> 00:10:38,200
It's not only a field level sensor. It's about also tracking the vehicles, also tracking

105
00:10:38,200 --> 00:10:44,980
the whole operations, so weighing systems on trucks, cameras. So basically we decided

106
00:10:44,980 --> 00:10:51,000
to have a hardware agnostic software, which can also create value without the hardware.

107
00:10:51,000 --> 00:10:57,280
We were actually trying to solve our own problems because being an IoT company, being an hardware

108
00:10:57,280 --> 00:11:02,920
company, also brings challenges because it's actually problematic from the supply chain

109
00:11:02,920 --> 00:11:10,440
point of view. It's also capital intensive. In order to make sales, you need to make expandings.

110
00:11:10,440 --> 00:11:15,220
Then you can make money afterwards. So it's also a challenge from the financial point.

111
00:11:15,220 --> 00:11:20,860
We said, hey, let's try to simplify things. So let's try to sell software. We also were

112
00:11:20,860 --> 00:11:25,720
not aware of software as a service concept. So we were actually trying to solve our own

113
00:11:25,720 --> 00:11:31,040
problem this time because there was a kind of bottleneck. Then you're getting sensor

114
00:11:31,040 --> 00:11:37,380
data and sometimes you're getting wrong data. Sometimes there are problems on the containers.

115
00:11:37,380 --> 00:11:44,160
We were also getting the temperature data from the sensors because we would like to

116
00:11:44,160 --> 00:11:50,820
understand if there is any potential fire inside of the bin, which is common in some

117
00:11:50,820 --> 00:11:56,720
areas. Then we said, okay, we need to understand the anomalies without checking manually. So

118
00:11:56,720 --> 00:12:03,140
basically we created an anomaly detection algorithm in order to understand if there's a potential

119
00:12:03,140 --> 00:12:10,500
fire inside. And then also we developed an algorithm to predict when bin is going to

120
00:12:10,500 --> 00:12:16,140
be full because sometimes bin is not full right now. If I'm not collecting this bin

121
00:12:16,140 --> 00:12:22,860
today, it is going to be overfull till next root. So I have to put this bin into my root

122
00:12:22,860 --> 00:12:27,900
plan. So I need to have the prediction. So basically these kinds of things are forced

123
00:12:27,900 --> 00:12:33,020
by the problems and we were trying to solve them. Then, then all of a sudden, of course,

124
00:12:33,020 --> 00:12:38,140
it's, it's not happening in one day, but we have the anomaly detection. We have the prediction.

125
00:12:38,140 --> 00:12:45,940
We have all these decision making processes. We have developed them all internally. Fantastic.

126
00:12:45,940 --> 00:12:52,600
So AI was that a topic back then. So basically to solve your own problems, you had to touch

127
00:12:52,600 --> 00:12:58,420
upon all the trending topics of the past 10 years. Well, I think, I think this is more,

128
00:12:58,420 --> 00:13:05,140
you know, this satisfies our internal, you know, the motivation a lot more instead of,

129
00:13:05,140 --> 00:13:12,480
you know, running after the buzzwords to be doing like real things in our mind. So because

130
00:13:12,480 --> 00:13:18,180
we see the problem, we decided to solve it. We solve it. I have it different algorithms

131
00:13:18,180 --> 00:13:24,340
or different approaches that we can think of. And well, it becomes IoT, it becomes SASS,

132
00:13:24,340 --> 00:13:30,340
it becomes machine learning, it becomes AI at the end, because we believe we shouldn't

133
00:13:30,340 --> 00:13:37,700
put those buzzwords as aim. We should utilize them for good purpose. Our good purpose is

134
00:13:37,700 --> 00:13:42,940
to solve the problem of waste and material management all around the earth. As much as

135
00:13:42,940 --> 00:13:49,640
we see problems, we need to solve it. So basically if it is AI today, then obviously we can implement

136
00:13:49,640 --> 00:13:56,340
AI, but it is not to implement AI. It is to solve problems, we can implement AI. Very

137
00:13:56,340 --> 00:14:01,740
beautifully said. I think I see that it could become a clip for us because a lot of, I see

138
00:14:01,740 --> 00:14:07,800
that a lot on LinkedIn and X that, hey, they're just like AI something. But why do you want

139
00:14:07,800 --> 00:14:14,100
to AI it? Do we actually need it? And, you know, touching upon my own personal experience

140
00:14:14,100 --> 00:14:19,880
with AI, and I've been working on multiple projects for the past months and I always wanted

141
00:14:19,880 --> 00:14:26,000
to add AI to it, but I came up with a very simple problem that need a very smart table.

142
00:14:26,000 --> 00:14:32,180
And there is no AI in it. It actually could solve one of my own problems that I have on

143
00:14:32,180 --> 00:14:38,380
a daily basis. So I decided to build it with AI, but there is no AI in it. True. True.

144
00:14:38,380 --> 00:14:43,140
Yeah. Sometimes it's just, you know, trying to kill a mosquito with a rocket instead of,

145
00:14:43,140 --> 00:14:49,000
you know, simple solutions. Yeah. Yeah. Really, I really tried to focus on simpler approaches

146
00:14:49,000 --> 00:14:55,160
instead of making it more complex or just to take it with a fancy things, but maybe this

147
00:14:55,160 --> 00:15:00,820
is because we are coming from Ankara approach. Maybe this is not correct. I don't know. This

148
00:15:00,820 --> 00:15:05,520
is our approach. I'm telling this is our approach. I'm not telling this is the best approach.

149
00:15:05,520 --> 00:15:11,880
This is our story. I love it. I love it. Being problem obsessed. I think that guarantees that,

150
00:15:11,880 --> 00:15:16,920
that you end up building technologies that you can reuse a lot. And then if I want to

151
00:15:16,920 --> 00:15:22,920
like lay out what you have done the past 10 years, now actually you're going to one of

152
00:15:22,920 --> 00:15:30,960
the very, very few companies that going to unlock so much potential from LLMs and GeneAiR.

153
00:15:30,960 --> 00:15:37,380
Because at the root of your platform, there is a IoT, there are machine learning algorithms

154
00:15:37,380 --> 00:15:42,840
that could detect and predict and there's a SaaS on top of it that offers all of this.

155
00:15:42,840 --> 00:15:50,640
Then it's a perfect transition toward perfect evolution towards using LLMs and GeneAi for

156
00:15:50,640 --> 00:15:58,840
offering more services and products. True. So hopefully we will by the help of LLMs and

157
00:15:58,840 --> 00:16:07,840
AI, we will be solving the problem from the deeper areas and not like from the layer,

158
00:16:07,840 --> 00:16:14,520
first layer only. Because if you try to solve the bigger problem, then you need to cover

159
00:16:14,520 --> 00:16:20,760
different stakeholders. One more stakeholder should be involved. There are different stakeholders,

160
00:16:20,760 --> 00:16:27,360
like the driver of the vehicle, the dispatcher, the management of the waste management company,

161
00:16:27,360 --> 00:16:33,600
security authorities and individuals like yourself, myself. So we are generating waste in different

162
00:16:33,600 --> 00:16:40,400
areas and so the companies, or producers, manufacturers, extended producer responsibility. So everyone

163
00:16:40,400 --> 00:16:46,000
is going to be involved. Then the problem is also getting bigger. Who is going to do

164
00:16:46,000 --> 00:16:53,200
what? Am I actually segregating my waste properly so that the waste management company can do

165
00:16:53,200 --> 00:16:59,120
the best of the operation? Because if I don't segregate properly, I am actually kind of

166
00:16:59,120 --> 00:17:05,840
creating another problem for the waste management companies. There is a need for increasing

167
00:17:05,840 --> 00:17:12,520
my awareness or changing my behavior to do good. It's not only the waste collection or

168
00:17:12,520 --> 00:17:21,240
waste treatment. It's also enabling waste generators to be part of the solution. We also

169
00:17:21,240 --> 00:17:29,100
try to involve waste generators in different perspectives. Let's move towards the visual

170
00:17:29,100 --> 00:17:37,480
side of our podcast. This is going to be recorded and the video will be available on Spotify

171
00:17:37,480 --> 00:17:45,740
and YouTube. We started with IoT. We need sensors to understand what's going on in this

172
00:17:45,740 --> 00:17:54,760
ecosystem to get that out of the need to detect and predict. Then we need basically a window

173
00:17:54,760 --> 00:18:02,680
of software that we can offer these services to our customers. What kind of problems that

174
00:18:02,680 --> 00:18:11,400
you've been facing that you couldn't solve with this tech stack that you then saw a lot

175
00:18:11,400 --> 00:18:21,080
of potential to solve with LLMs and GNI? Could you ask it again? Okay. Up to LLMs, your tech

176
00:18:21,080 --> 00:18:30,960
stack was IoT sensors of different kinds. You got SaaS to offer your services. You got

177
00:18:30,960 --> 00:18:37,960
deep learning to detect and predict. Then now LLM came to the playground. What kind

178
00:18:37,960 --> 00:18:44,600
of problem you've been facing that you couldn't solve with your previous tech stack now and

179
00:18:44,600 --> 00:18:52,320
then with LLM you could solve? Yeah, of course. Basically, we have started as an IoT company,

180
00:18:52,320 --> 00:18:59,080
and then we decided to become a SaaS company. That was the second step. Then we realized

181
00:18:59,080 --> 00:19:04,680
that if you need to solve problems of different countries, different waste streams, different

182
00:19:04,680 --> 00:19:11,120
regulations, then your solution should be flexible enough to cover all the municipal waste operation

183
00:19:11,120 --> 00:19:17,660
in the Netherlands and also electronic waste in Singapore and medical waste in South Africa.

184
00:19:17,660 --> 00:19:23,240
Different regulations, different operations, everything. Currently, we are using the third

185
00:19:23,240 --> 00:19:30,640
version of our product stack. This is actually the rule set mechanism, because our customers

186
00:19:30,640 --> 00:19:37,560
can create their own if this, then that structures. So, we actually give them this capability

187
00:19:37,560 --> 00:19:44,880
to create their own rule sets. By this way, we are capable to onboard our customers in

188
00:19:44,880 --> 00:19:51,380
weeks time. So, we don't need to change the product from scratch or we don't need to customize

189
00:19:51,380 --> 00:19:58,160
our product. We just configure the environment and give them the flexibility to change their

190
00:19:58,160 --> 00:20:06,920
own rule sets. So, the same solution can be used in different areas for different purposes.

191
00:20:06,920 --> 00:20:14,640
So, that is one of the good things. That was a really good approach, which is kind of unique

192
00:20:14,640 --> 00:20:20,760
in the market. So, this is actually differentiating our solution from the other providers, and

193
00:20:20,760 --> 00:20:26,360
there was a kind of challenge related with the dashboard part of it. Because one company

194
00:20:26,360 --> 00:20:31,000
would like to see different KPIs and the other company would like to see different KPIs.

195
00:20:31,000 --> 00:20:36,200
For that purpose, we need to understand what kind of KPIs they have and we need to build

196
00:20:36,200 --> 00:20:44,240
all these dashboard things on our own. But obviously, those KPI requirements are changing

197
00:20:44,240 --> 00:20:50,080
in time. So, maybe they wanted these 10 KPIs from the first day, two months later, they

198
00:20:50,080 --> 00:20:53,720
would like to see another KPI. That's right. Three months later, they would like to see

199
00:20:53,720 --> 00:20:59,840
another two. So, this is actually lots of requirements coming from the customers. We try to give

200
00:20:59,840 --> 00:21:04,120
response to them with our customer success, product management, product development, it's

201
00:21:04,120 --> 00:21:11,140
actually consuming a lot. So, we were spending too much of time in order to satisfy these

202
00:21:11,140 --> 00:21:21,600
requirements. But thanks to LLM, we gave the capability to our customers to ask questions

203
00:21:21,600 --> 00:21:29,780
to their data. So, they can basically ask which vehicle is performing best, which driver

204
00:21:29,780 --> 00:21:36,080
is doing good, which driver is just, you know, not doing good. So, basically, data set is

205
00:21:36,080 --> 00:21:43,240
there although the dashboard is kind of, you know, you have 10 KPIs, but you can ask questions

206
00:21:43,240 --> 00:21:51,560
to your dashboard. So, we don't get requirements like 12th, 13th KPI, we gave them the capability

207
00:21:51,560 --> 00:21:57,240
to communicate with their own tool so that they can gather any data, any KPIs from the

208
00:21:57,240 --> 00:22:02,400
system on their own. Obviously, we are developing and improving this capability, but yeah, this

209
00:22:02,400 --> 00:22:08,680
is available before it. Now, it is available thanks to it. So, we also secured some of

210
00:22:08,680 --> 00:22:10,600
our time and attention.

211
00:22:10,600 --> 00:22:17,600
Can you map the, so by implementing this, by implementing LLMs, to allow your users

212
00:22:17,600 --> 00:22:24,960
to go layers deeper than what they see on the interface and interact with the data sets?

213
00:22:24,960 --> 00:22:32,480
Most likely, the byproduct of that, the causation of, well, correlation of a relationship should

214
00:22:32,480 --> 00:22:40,040
be that it did reduce the burden that is on your customer service and on your product

215
00:22:40,040 --> 00:22:42,800
team, support team, whatever the case may be. Is that correct?

216
00:22:42,800 --> 00:22:48,120
Yeah, true, true, definitely. Because if you give them the capability to gather what they

217
00:22:48,120 --> 00:22:54,560
want from the solution. So, instead of creating a ticket for our customer success team, they

218
00:22:54,560 --> 00:23:00,680
can actually get the response because actually they communicate with their own words. So,

219
00:23:00,680 --> 00:23:06,440
normally what they do, they write words to our customer success team, "Hey, I would like

220
00:23:06,440 --> 00:23:13,240
to see this data." And actually they are writing these messages to their system, so the system

221
00:23:13,240 --> 00:23:19,800
is actually replying back directly, so they can see those data on their own instead of

222
00:23:19,800 --> 00:23:26,320
creating tickets. So, this is actually helping them to gather the data that they would like

223
00:23:26,320 --> 00:23:32,420
right away. And this is actually reducing, you know, the effort that our customer success

224
00:23:32,420 --> 00:23:36,480
team and product teams spend to resolve these tickets.

225
00:23:36,480 --> 00:23:43,680
Beautiful, beautiful. I think that's a huge advantage of LLM, because typically I think

226
00:23:43,680 --> 00:23:51,080
one of the frustrations that none of us in the design and product could solve prior to

227
00:23:51,080 --> 00:23:57,040
LLM when we were designing SaaS applications is that, for example, typical SaaS application

228
00:23:57,040 --> 00:24:03,120
has a reporting page, insight page, home page, whatever the, how you want to name it. And

229
00:24:03,120 --> 00:24:08,560
you would go there and to your point, you would see a set of KPIs. At most, there was

230
00:24:08,560 --> 00:24:12,900
a filter or there was something configuration with the settings that we could say that,

231
00:24:12,900 --> 00:24:19,280
"Okay, instead of seeing these KPIs, I want to see these KPIs." But basically, you had

232
00:24:19,280 --> 00:24:26,840
to understand or you had to come to this realization by yourself through going through the data,

233
00:24:26,840 --> 00:24:33,080
through the data, to realize that, okay, instead of having this set of KPIs, I want to use

234
00:24:33,080 --> 00:24:40,640
this set of KPIs, right? So with LLM, you can embark on this journey of, "Okay, let's

235
00:24:40,640 --> 00:24:48,480
get down under the hood and go layers deep to get a better sense of my data and how,

236
00:24:48,480 --> 00:24:53,060
for example, my customers are doing." So you can embark on that journey.

237
00:24:53,060 --> 00:25:00,400
So my question here, you said that with LLM, users or customers can now understand far

238
00:25:00,400 --> 00:25:06,000
better what's going on with their customers and users and have access to a broader range

239
00:25:06,000 --> 00:25:12,000
of their data set. When they fetch the information that is typically not available under SaaS

240
00:25:12,000 --> 00:25:20,280
dashboard, can they with prompt or perhaps with some interface configuration change the

241
00:25:20,280 --> 00:25:26,660
dashboard the way they want it to see it? Currently, that is not available in our product.

242
00:25:26,660 --> 00:25:31,640
So there are a couple of things which are in our roadmap. One of the things is this,

243
00:25:31,640 --> 00:25:39,240
because having a conversation with your system and getting the response is the first thing,

244
00:25:39,240 --> 00:25:45,840
but then I would like to implement what I have learned, right? That is going to be the

245
00:25:45,840 --> 00:25:52,120
second thing, because I'm learning it. So what? Then I'm going to write it down, then I'm

246
00:25:52,120 --> 00:25:57,240
going to do it manually. All of these steps, again, needs to be done. So this is one of

247
00:25:57,240 --> 00:26:02,360
the things, because you can ask questions, you can brainstorm with it, and then you need

248
00:26:02,360 --> 00:26:06,960
to make it implement. So we are not there yet. So it's like step-by-step, again, the

249
00:26:06,960 --> 00:26:12,220
same approach. So that was the first problem. Then what? Then this is the second problem.

250
00:26:12,220 --> 00:26:17,480
So we have to solve that one too. So this is one of the things which is in our roadmap.

251
00:26:17,480 --> 00:26:23,140
The other thing, because there are, I mentioned two stakeholders, like waste collection, waste

252
00:26:23,140 --> 00:26:29,600
treatment, and the third in general, waste generators. We also have waste generator customers.

253
00:26:29,600 --> 00:26:36,040
Waste generators can be individuals or big companies, like manufacturing facilities,

254
00:26:36,040 --> 00:26:43,980
imagine an automotive facility, building tens of thousands of vehicles. Then they are also

255
00:26:43,980 --> 00:26:49,360
generating huge amount of waste from different types of materials, in hazardous, non-hazardous,

256
00:26:49,360 --> 00:26:55,680
so on, so forth. So currently they are at point A, but actually without our solution,

257
00:26:55,680 --> 00:27:03,800
they cannot describe the point A. They don't know how much waste they are generating, because

258
00:27:03,800 --> 00:27:11,880
of which activity. I see. At which corner of their facility, in which shift, which type

259
00:27:11,880 --> 00:27:17,880
of material, hazardous, non-hazardous, are they doing good? Are they segregating the

260
00:27:17,880 --> 00:27:23,040
materials properly, or are they mixing hazardous and non-hazardous material, which is going

261
00:27:23,040 --> 00:27:28,360
to be hazardous at the end? So these kinds of things. The only data that they have is

262
00:27:28,360 --> 00:27:33,720
the invoices of the waste management companies. It's like ballpark figure at the end of the

263
00:27:33,720 --> 00:27:39,800
month or quarter. So if you don't know the details of all these things that I mentioned,

264
00:27:39,800 --> 00:27:45,240
you cannot improve your performance from the waste management point of view as a waste

265
00:27:45,240 --> 00:27:50,700
generator. Right. Which is like, you know, doctor is looking at you. Hey, you look okay.

266
00:27:50,700 --> 00:27:57,600
Right. They do blood test, right? Because maybe in your blood, there are different numbers,

267
00:27:57,600 --> 00:28:04,320
some like less, some higher. So looking at those numbers, they can make a decision. So

268
00:28:04,320 --> 00:28:10,180
it is because the waste or urine as well. Urine test as well. Waste is kind of blood test

269
00:28:10,180 --> 00:28:18,680
or urine test. Imagine you make a... It's such a life's analogy. Yeah. Basically it's like,

270
00:28:18,680 --> 00:28:24,480
you know, continuous testing your body. And if there is something happening, you can understand

271
00:28:24,480 --> 00:28:31,460
from the waste. Yeah. Yeah. Makes sense to me. We are again in anomaly detection. Yeah.

272
00:28:31,460 --> 00:28:36,600
That is kind of thing. And coming to the point A, they don't know point A. They cannot describe

273
00:28:36,600 --> 00:28:41,220
where they are right now. And actually by the regulation, by the company policies, they

274
00:28:41,220 --> 00:28:48,980
have targets maybe like 10, 15 years. Let's say that is point Z. They can define like

275
00:28:48,980 --> 00:28:54,600
ballpark figure point Z. They don't know where they are right now. Point ABC. So what our

276
00:28:54,600 --> 00:29:02,720
aim is to first describe point A, describe point Z and create a roadmap from point A

277
00:29:02,720 --> 00:29:09,480
to Z in 10 years, 15 years. It's not only roadmap. This is like targets, but how can

278
00:29:09,480 --> 00:29:15,400
I achieve those targets? What are my action points? What should I do to achieve my target

279
00:29:15,400 --> 00:29:23,080
within three months? So these are actually possible with AI. It is a little complex problem.

280
00:29:23,080 --> 00:29:29,000
I know, but this is doable. Now we are working on these two. One is putting the output of

281
00:29:29,000 --> 00:29:35,760
our, you know, brainstorming into action. Two, creating a roadmap and making sure that

282
00:29:35,760 --> 00:29:45,580
I can achieve my long term target by forcing, you know, action points to my field employees,

283
00:29:45,580 --> 00:29:49,120
because they don't know what to do tomorrow in the morning. We need to give them some

284
00:29:49,120 --> 00:29:55,120
actions. So yeah, these are the things. So basically first step is diagnosis. Yeah. What's

285
00:29:55,120 --> 00:30:02,040
wrong? Not wrong. No, what is the current status? Exactly. Just take a picture today.

286
00:30:02,040 --> 00:30:06,740
What's the, what's the current, what's, what's in your blood? So waste blood makes sense

287
00:30:06,740 --> 00:30:13,140
to me all the way. Then the second point with AI will be progresses. Okay, now this is your

288
00:30:13,140 --> 00:30:19,600
current state of effort. This is the goal is set by regulatory bodies. You have to achieve

289
00:30:19,600 --> 00:30:25,140
it whether you like it or not within 15 years. Let us generate you a roadmap. The roadmap

290
00:30:25,140 --> 00:30:31,400
is based on the goal set by regulatory bodies and the best industry practices for waste

291
00:30:31,400 --> 00:30:38,000
management and treatment and collection. Yep. True. Right. And obviously it has to be reasonable

292
00:30:38,000 --> 00:30:42,960
for the organization. That's a really good point too. It cannot be like sort of a blue

293
00:30:42,960 --> 00:30:49,160
sky goal that you have to next, next year you have to reduce waste by 50%. Yeah, it's,

294
00:30:49,160 --> 00:30:55,360
it's not going to happen. That is why, you know, the waste is not core thing for any

295
00:30:55,360 --> 00:31:00,200
organization other than the, you know, waste management companies. Imagine you, you are

296
00:31:00,200 --> 00:31:06,800
a car manufacturing facility. Your main objective is to, you know, building vehicles, making

297
00:31:06,800 --> 00:31:12,560
profit. Managing the waste is like, I have to do it some way and I have to, you know,

298
00:31:12,560 --> 00:31:19,000
comply. That's it. So it's not your main objective. So it has to be handled seamlessly. It's,

299
00:31:19,000 --> 00:31:25,280
it shouldn't be a big problem for my operation. It shouldn't stop me from making enough production,

300
00:31:25,280 --> 00:31:32,140
making enough profit, but it has to be done somehow. So it has to be achievable, reasonable.

301
00:31:32,140 --> 00:31:37,680
I guess for those kinds of negotiations, you would wear your Istanbul hat and try to connect

302
00:31:37,680 --> 00:31:43,680
how waste management could improve your profit at the end. Definitely true. But by the way,

303
00:31:43,680 --> 00:31:50,500
it actually has an impact because waste, managing the waste has an impact on cost or has an

304
00:31:50,500 --> 00:31:56,680
impact on revenue. Because if you have valuable material, this can be a kind of revenue. Obviously,

305
00:31:56,680 --> 00:32:03,180
it's not as much as selling the products, right? But it can be a certain amount of cost or

306
00:32:03,180 --> 00:32:08,440
revenue depending on how you manage your material. And also could be a very good use case for

307
00:32:08,440 --> 00:32:14,760
LLM too. If you say that, okay, we can save you this amount of waste and basically calculate

308
00:32:14,760 --> 00:32:21,040
at how much that base reduction would lead to reducing costs across the board in your

309
00:32:21,040 --> 00:32:27,640
operation and increasing profit. I think any CFO would love to see that chart. Yeah, definitely.

310
00:32:27,640 --> 00:32:33,760
Definitely. So we would like to convince them first, so that because generally, you know,

311
00:32:33,760 --> 00:32:40,180
these targets to whole organization and then goes to directors, goes to managers, goes

312
00:32:40,180 --> 00:32:45,440
to all the employees. Obviously, it is coming from top in general. So if we can convince

313
00:32:45,440 --> 00:32:52,880
them from their financial drive forces, then it is going to be spread to all organizations.

314
00:32:52,880 --> 00:32:58,600
It is not enough to convince the environmental manager within the organization only because

315
00:32:58,600 --> 00:33:05,280
it's going to be in his or her area solely, which is not going to have an impact on the

316
00:33:05,280 --> 00:33:12,520
operation side. But if organization is convinced from the top, then everything can be possible.

317
00:33:12,520 --> 00:33:17,320
Of course, there will be some challenges in that case too, because it's going to be a

318
00:33:17,320 --> 00:33:23,640
big change management problem. But still, it is better to convince most of the different

319
00:33:23,640 --> 00:33:28,480
stakeholders in different layers. Mujer, I think it's a really good time to jump into

320
00:33:28,480 --> 00:33:35,080
the interface and see what are you being cooking for the past 10 years? Of course. Just give

321
00:33:35,080 --> 00:33:40,400
me a second so I can... I'm really excited now after like 40 minutes of going down the

322
00:33:40,400 --> 00:33:46,800
rabbit hole of waste. So basically, I would like to show you a couple of pages. Obviously,

323
00:33:46,800 --> 00:33:52,520
not all. Let me share my screen. If you're listening to these podcasts throughout your

324
00:33:52,520 --> 00:34:02,760
channels, you can easily get the timestamp, go to YouTube or Spotify and see the screens.

325
00:34:02,760 --> 00:34:08,280
Can you see my screen now? Yes. Perfect. So basically, this is like a master demo account

326
00:34:08,280 --> 00:34:18,200
of our dashboard. Obviously, this data is just randomly generated data. So this is like

327
00:34:18,200 --> 00:34:28,120
the main dashboard. All these boxes can be configured depending on different KPI requirements,

328
00:34:28,120 --> 00:34:34,320
as we have taught. So we were generating these boxes per customer because they would like

329
00:34:34,320 --> 00:34:42,640
to see different types of KPI. And we have different modules like operation, we have

330
00:34:42,640 --> 00:34:49,240
fleet management, we have assets. Assets is like the bins and containers. We have engagement.

331
00:34:49,240 --> 00:34:56,280
Engagement is actually related with the relationship management for waste management companies,

332
00:34:56,280 --> 00:35:02,080
so they can manage other third parties here like subcontractors, their customers, authorities.

333
00:35:02,080 --> 00:35:08,920
MRF is for the material recovery facility, so they can manage incoming materials, processes

334
00:35:08,920 --> 00:35:14,960
within the facility, inventory management, and the outgoing materials to their different

335
00:35:14,960 --> 00:35:19,800
types of customers. An employee is managing the employees because different employees

336
00:35:19,800 --> 00:35:29,920
has different constraints. So basically, they can have the drivers. That driver can only

337
00:35:29,920 --> 00:35:35,440
use this type of vehicle, maybe the employee which is working on field only. They cannot

338
00:35:35,440 --> 00:35:40,440
drive any vehicle. So they have different constraints too. And we have user management

339
00:35:40,440 --> 00:35:47,520
events, which is like the reports coming from the field, complaints, et cetera. And all

340
00:35:47,520 --> 00:35:56,060
the reports can be also archived here. So anyone who would like to see those reports

341
00:35:56,060 --> 00:36:01,400
afterwards, they can go to this archive and see all the reports. And there are different

342
00:36:01,400 --> 00:36:07,800
document types here. And our system is also supporting different languages. Currently,

343
00:36:07,800 --> 00:36:15,520
12 languages we support, which is actually easy to translate to different languages.

344
00:36:15,520 --> 00:36:20,640
And if you want to see the one month data, three months data, six months data, these

345
00:36:20,640 --> 00:36:27,720
are the big challenges. So what I would like to show you just simply is the AI button here.

346
00:36:27,720 --> 00:36:35,960
So basically, either you can ask these kind of questions, predefined questions, go for

347
00:36:35,960 --> 00:36:42,760
the vehicle performance, then I can see the vehicle performance data here, then I can

348
00:36:42,760 --> 00:36:53,640
ask questions here, or I can directly communicate with sustainability assistant here. So I can

349
00:36:53,640 --> 00:37:04,380
ask questions to my chat bot. So how can I reduce, how to say, my recycling rate, increase

350
00:37:04,380 --> 00:37:11,720
my recycling rate within my city, then I can even get responses to that. So this is basically

351
00:37:11,720 --> 00:37:18,160
what we are showing to our customers. But not all of them are enjoying right now because

352
00:37:18,160 --> 00:37:25,740
this is like special trial for some of them. So after we can be satisfied with the old

353
00:37:25,740 --> 00:37:30,480
results, then we are going to be launching to all the customers. So this is what we are

354
00:37:30,480 --> 00:37:39,240
currently testing with some of the customers. So for example, in that anomaly analysis,

355
00:37:39,240 --> 00:37:44,560
basically, whenever I go there, if there is an anomaly with my operation, I could see it.

356
00:37:44,560 --> 00:37:50,480
So anomaly detection is one related with the fill level sensors, because in fill level

357
00:37:50,480 --> 00:37:57,000
sensors, there might be two main challenges. One is the potential fire, because if there

358
00:37:57,000 --> 00:38:04,040
is, you know, the high temperature doesn't mean a fire, but increasing rapidly may mean

359
00:38:04,040 --> 00:38:11,740
a fire, for instance. So this is different because we work in the very hot geographies

360
00:38:11,740 --> 00:38:18,000
like Saudi UAE area. So the temperature within the metal container can be really high, but

361
00:38:18,000 --> 00:38:22,460
this doesn't mean a fire. So this is one of the anomalies that we can detect. The other

362
00:38:22,460 --> 00:38:29,640
one is the data, like fill level data can be problematic as well. Maybe it is, you know,

363
00:38:29,640 --> 00:38:34,460
there is a blockage in front of the sensor. Maybe there is a problem with the sensor.

364
00:38:34,460 --> 00:38:39,280
So those kinds of anomalies can be detected too. So this is actually going to like the

365
00:38:39,280 --> 00:38:45,680
fire is actually creating, remember the events. Yeah, this is actually creating an event alarm,

366
00:38:45,680 --> 00:38:52,040
saying that potential fire. So those events is actually created by the alarm mechanisms

367
00:38:52,040 --> 00:38:57,640
and anomaly detection for the sensors, which is actually our problem. It comes to our customer

368
00:38:57,640 --> 00:39:03,200
success team. There might be a problem within your sensor. Just check it manually. And if

369
00:39:03,200 --> 00:39:08,840
you confirm that there is a problem with the sensor, then try to solve it. If you cannot

370
00:39:08,840 --> 00:39:14,440
solve it, you need to maybe change the sensor whatsoever. Before this anomaly detection,

371
00:39:14,440 --> 00:39:20,840
our customer success managers had to check the sensors in a frequent manner. Like every

372
00:39:20,840 --> 00:39:27,360
week they had a dedicated time to check the sensors one by one. Then we said, "Okay, let's

373
00:39:27,360 --> 00:39:33,160
create an algorithm because after some time, we will not be able to check all of them manually."

374
00:39:33,160 --> 00:39:40,680
So that was the starting point. But the third, it's not anomaly detection, but it is actually

375
00:39:40,680 --> 00:39:46,640
performance management. Recently we launched it. Imagine you have a hundred tasks in your

376
00:39:46,640 --> 00:39:53,260
route, right? And your route length is eight hours. Like four hours later than you started.

377
00:39:53,260 --> 00:40:01,440
You expect to finalize 50 of those tasks. If you finalized so far, if you finalize 30

378
00:40:01,440 --> 00:40:06,720
of them, you're actually, you know, behind the expected schedule. And we are predicting

379
00:40:06,720 --> 00:40:13,000
that you are not going to be able to finalize all of the tasks of your routes. Then we do

380
00:40:13,000 --> 00:40:19,480
the calculation. How many tasks is going to be undone till the end of the day. Then we

381
00:40:19,480 --> 00:40:25,280
recommend the dispatcher, "Hey, there is a problem here in this route. Make an action.

382
00:40:25,280 --> 00:40:33,520
Take those potentially risky tasks and assign it to another route." By this way we can assure

383
00:40:33,520 --> 00:40:40,240
at the end of the route, at the end of the day, we are maximizing the performance of

384
00:40:40,240 --> 00:40:45,800
all the routes. Otherwise we can only show them, you know, "Hey, you have done 80% today.

385
00:40:45,800 --> 00:40:50,700
You have done 70% today. You have done 90% today." This is actually showing what has

386
00:40:50,700 --> 00:40:57,400
happened. This is also valuable, but our aim is to enable our customers to maximize their

387
00:40:57,400 --> 00:41:02,160
performance. It's not only show them how they perform. So instead of like them hitting their

388
00:41:02,160 --> 00:41:11,600
quota 80%, let's focus on creating performance impact outcome. Yeah. Would it be, so basically

389
00:41:11,600 --> 00:41:18,580
the way you are thinking about it, it's better off for a vehicle to hit 60% what covered

390
00:41:18,580 --> 00:41:26,580
the most riskiest cases in the route, rather than hitting 80%, but not covering those.

391
00:41:26,580 --> 00:41:31,980
Yes, definitely. So I can, it is not in my master demo account available, but I can send

392
00:41:31,980 --> 00:41:37,080
you a screenshot if you want. You can put them as well. You can share them too, because

393
00:41:37,080 --> 00:41:43,820
you actually put them in a, you know, red route layer, like on top. So the dispatcher is actually

394
00:41:43,820 --> 00:41:50,400
checking them and managing them, because I don't want to say, "Hey, Behrat, you didn't

395
00:41:50,400 --> 00:41:56,280
manage your tasks today." It's actually nothing valuable for me as a dispatcher. My aim is

396
00:41:56,280 --> 00:42:03,100
to maximize the fulfillment rate of my total routes. Right, right. It's not about who is

397
00:42:03,100 --> 00:42:09,780
performing how, it's about fulfilling all the tasks. So currently, we are suggesting to

398
00:42:09,780 --> 00:42:16,880
dispatcher, but after learning their decision-making processes, we are going to assign these risky

399
00:42:16,880 --> 00:42:23,720
tasks to automated selected routes without asking to dispatcher. So we will be actually

400
00:42:23,720 --> 00:42:30,680
managing all these operations by the system. Omarjan, I think if we talk at the end of

401
00:42:30,680 --> 00:42:38,440
this year, you're probably going to break out how you guys were not a robotic company. Now

402
00:42:38,440 --> 00:42:45,760
you are a robotic company. I don't know. I don't know. Maybe we need to think about it

403
00:42:45,760 --> 00:42:50,920
as that, because this is actually how we are trying to solve. Normally, the dispatchers

404
00:42:50,920 --> 00:42:55,680
are checking the WhatsApp groups, trying to understand what is happening, calling drivers

405
00:42:55,680 --> 00:43:00,660
like all the time. So instead of getting pictures on WhatsApp groups and calling other drivers,

406
00:43:00,660 --> 00:43:07,280
and managing all of them like super chaotic way, we had them to manage like in a calm

407
00:43:07,280 --> 00:43:13,600
environment, let them think wisely, manage it with data, do it afterwards. But actually,

408
00:43:13,600 --> 00:43:19,220
our system can learn from their activities so that they the system can implement at the

409
00:43:19,220 --> 00:43:25,800
end. But there is also this question comes, who is the responsible of this route management?

410
00:43:25,800 --> 00:43:32,800
Is it our system or is it the dispatcher itself? So maybe we will just ask, just confirm button.

411
00:43:32,800 --> 00:43:40,120
YOLO mode. You know, exactly. You only look once. You are getting to the YOLO mode sooner

412
00:43:40,120 --> 00:43:45,500
or later. And I think it's better off to for this kind of really like complex route management.

413
00:43:45,500 --> 00:43:51,320
I mean, you don't have to rely on YOLO mode all the time, but having a YOLO mode is better

414
00:43:51,320 --> 00:43:54,640
than not having a YOLO mode. I think so too. Yeah. Otherwise, it can be a little bit complicated.

415
00:43:54,640 --> 00:43:59,520
Otherwise, it can be a little challenging. So from the from the optimization point of

416
00:43:59,520 --> 00:44:08,520
view, it is also about estimated time of arrival. Yeah. Because some of the customers have time

417
00:44:08,520 --> 00:44:13,700
window, some of the customers of our customers have some time windows, they only say, hey,

418
00:44:13,700 --> 00:44:21,580
you can pick my waste from 9 to 10. So in order to make this route optimization, we

419
00:44:21,580 --> 00:44:26,620
need to take this into consideration as well. So we do the route optimization, like the

420
00:44:26,620 --> 00:44:33,780
sequencing, but we need to be sure that that time window should be satisfied too. So these

421
00:44:33,780 --> 00:44:41,820
kind of challenges are also coming and going. And obviously, we there are still some challenges

422
00:44:41,820 --> 00:44:49,700
that we are not able to solve, especially related with the map data. If there is no relevant

423
00:44:49,700 --> 00:44:56,080
map data, then you cannot manage whatever you do. So you need to have a proper real time

424
00:44:56,080 --> 00:45:02,240
map data as well. So many, so many variables and constraints. I got to say, I thoroughly

425
00:45:02,240 --> 00:45:09,480
enjoy this conversation as an engineer, as a UX designer, it has been a very hands-on

426
00:45:09,480 --> 00:45:16,720
conversation with a founder, a visionary person who's very obviously clear that you're like

427
00:45:16,720 --> 00:45:22,820
dealing with this hard problem on a daily basis, minute basis, hour basis. For me, it has been

428
00:45:22,820 --> 00:45:29,320
so refreshing to understand that what it takes, I mean, I just like, like listen to it for

429
00:45:29,320 --> 00:45:36,600
15 minutes, but the kind of hard problems you have to solve, and it's physics, some of them

430
00:45:36,600 --> 00:45:41,600
are really physics. You have to deal with physics, that is physics is very unforgiving,

431
00:45:41,600 --> 00:45:46,340
you have to, it doesn't matter if you have good will or not, it's very unforgiving. I

432
00:45:46,340 --> 00:45:52,240
don't know. I'm very impressed. I'm very wow. Thank you for being very like, obviously,

433
00:45:52,240 --> 00:45:56,460
this is, this is a business that you've been working on for 10 years, so everything is

434
00:45:56,460 --> 00:46:01,360
here. But thank you for bringing all this here because I think one thing that our listeners

435
00:46:01,360 --> 00:46:08,600
could get out with is that building meaningful businesses that solve hard problems is absolutely

436
00:46:08,600 --> 00:46:14,920
not a walk in the park. Yeah, yeah, definitely, definitely. Because obviously looking from,

437
00:46:14,920 --> 00:46:21,520
outside, the challenges can be seen small, you can think that you can solve it easily.

438
00:46:21,520 --> 00:46:25,520
Well, we taught the same way, by the way, we didn't think that it's going to be a really

439
00:46:25,520 --> 00:46:29,240
big problem, we thought it is going to be a simple problem that we can solve easily.

440
00:46:29,240 --> 00:46:34,960
It is not. Then we accept that this is not going to be a short journey. It is going to

441
00:46:34,960 --> 00:46:40,240
be a really big journey from the, from, from every perspective, by the way, it's not only

442
00:46:40,240 --> 00:46:45,160
technical challenges, like commercial, you know, relationships and everything. But if

443
00:46:45,160 --> 00:46:51,660
you enjoy the challenges, if you enjoy the problems and trying to solve it, that is actually

444
00:46:51,660 --> 00:46:58,680
kind of, you know, giving you the feeling of walking in the park. Yeah, actually, you

445
00:46:58,680 --> 00:47:05,320
know, you actually enjoy it. I myself is like an endurance sports lover. So you enjoy suffering,

446
00:47:05,320 --> 00:47:12,600
actually. I couldn't find a way, I couldn't think of a better way of ending this conversation.

447
00:47:12,600 --> 00:47:17,320
Thanks a lot, Ubu Munchen. Thank you. Thank you, Behrad. It was a lovely conversation.

448
00:47:17,320 --> 00:47:22,640
Thank you for listening to UX for AI. Join us next week for more insightful conversations

449
00:47:22,640 --> 00:47:28,880
about the impact of artificial intelligence in development, design and user experience.

450
00:47:28,880 --> 00:47:30,880
[MUSIC]