1
00:00:15,316 --> 00:00:27,756
Speaker 1: Pushkin. I'm Jacob Goldstein and this is What's Your Problem,

2
00:00:27,956 --> 00:00:30,716
Speaker 1: the show where entrepreneurs and engineers talk about how they're

3
00:00:30,756 --> 00:00:33,556
Speaker 1: going to change the world once they solve a few problems.

4
00:00:34,516 --> 00:00:38,196
Speaker 1: My guest today is Anna Katrina d Letski. She's a

5
00:00:38,236 --> 00:00:41,716
Speaker 1: mechanical engineer who started her career at Apple, where she

6
00:00:41,796 --> 00:00:46,836
Speaker 1: worked on the Apple Watch that was responsible for we

7
00:00:46,956 --> 00:00:52,396
Speaker 1: call it packing the suitcase. So the industrial design team

8
00:00:52,436 --> 00:00:55,236
Speaker 1: makes beautiful surfaces of like what it's going to look

9
00:00:55,276 --> 00:00:58,716
Speaker 1: like on the outside, and the product design engineers figure

10
00:00:58,756 --> 00:01:01,676
Speaker 1: out how to get all the parts inside that suitcase.

11
00:01:01,836 --> 00:01:03,956
Speaker 1: If you will, you got to get all the parts in,

12
00:01:04,276 --> 00:01:07,516
Speaker 1: and then you know that product needs to be robust.

13
00:01:08,076 --> 00:01:10,556
Speaker 1: When Anna was working on the watch, she saw a

14
00:01:10,596 --> 00:01:13,716
Speaker 1: massive problem, not just with the watch or even just

15
00:01:13,756 --> 00:01:17,916
Speaker 1: with Apple, but with all of electronics manufacturing. The problem

16
00:01:17,996 --> 00:01:23,876
Speaker 1: is this electronics manufacturing is incredibly inefficient, incredibly wasteful. And

17
00:01:23,916 --> 00:01:27,556
Speaker 1: I'm not talking about consumers throwing devices away. I'm talking

18
00:01:27,556 --> 00:01:31,676
Speaker 1: about the manufacturing process itself, what happens before the product

19
00:01:31,716 --> 00:01:35,396
Speaker 1: even leaves the factory. So Anna left Apple and started

20
00:01:35,436 --> 00:01:38,436
Speaker 1: a company called Instrumental to try to solve that problem

21
00:01:39,516 --> 00:01:42,916
Speaker 1: in our conversation. Anna laid out the problem really clearly

22
00:01:43,076 --> 00:01:46,276
Speaker 1: when she explained what happens when a new device is

23
00:01:46,316 --> 00:01:49,196
Speaker 1: about to go into production. It starts with a sort

24
00:01:49,236 --> 00:01:52,116
Speaker 1: of practice build, when a few hundred or a few

25
00:01:52,156 --> 00:01:55,476
Speaker 1: thousand devices are assembled as a test. If you're part

26
00:01:55,516 --> 00:01:58,516
Speaker 1: of a team of engineers working in your cushy office

27
00:01:58,516 --> 00:02:01,516
Speaker 1: in Kupertino, this is the moment when you and your

28
00:02:01,516 --> 00:02:04,356
Speaker 1: team have to go and actually see this thing you've

29
00:02:04,396 --> 00:02:07,556
Speaker 1: designed being built. You'd all get on a flight and

30
00:02:07,636 --> 00:02:10,716
Speaker 1: you'd fly in earlatest China. It's where a lot of

31
00:02:10,716 --> 00:02:13,316
Speaker 1: stuff is built, although today maybe it's somewhere else in

32
00:02:13,356 --> 00:02:17,876
Speaker 1: Greater Asia. And then you land in a city, and

33
00:02:17,916 --> 00:02:20,436
Speaker 1: then you'd get into a factory van and you might

34
00:02:20,516 --> 00:02:22,596
Speaker 1: drive for an hour and a half or two hours

35
00:02:22,596 --> 00:02:25,876
Speaker 1: out into a very rural part of China, and that's

36
00:02:25,876 --> 00:02:29,636
Speaker 1: where you'd have a massive manufacturing center. And then we

37
00:02:29,676 --> 00:02:32,956
Speaker 1: would be there for about ten or fourteen days. What

38
00:02:33,156 --> 00:02:35,396
Speaker 1: is the point of you being there. I'm there trying

39
00:02:35,396 --> 00:02:40,596
Speaker 1: to find issues, find problems, find parts that aren't fitting,

40
00:02:40,796 --> 00:02:44,756
Speaker 1: find quality issues with parts. So like we expect there

41
00:02:44,756 --> 00:02:47,916
Speaker 1: to be problems and quality issues with the parts, Anna said,

42
00:02:48,076 --> 00:02:51,356
Speaker 1: she's not allowed to talk about specific preproduction problems they

43
00:02:51,396 --> 00:02:53,956
Speaker 1: had with the Apple Watch, but she did give me

44
00:02:53,996 --> 00:02:58,876
Speaker 1: a hypothetical source of problems the antenna. So, okay, picture

45
00:02:58,916 --> 00:03:01,956
Speaker 1: of factory with hundreds of workers packing all these tiny

46
00:03:01,996 --> 00:03:04,676
Speaker 1: little parts into the watch. One of those parts is

47
00:03:04,756 --> 00:03:07,876
Speaker 1: the antenna, this tiny little thing the sizes of like

48
00:03:07,996 --> 00:03:11,036
Speaker 1: a one inch piece of dried spaghetti, and it has

49
00:03:11,076 --> 00:03:13,036
Speaker 1: to go in there just right so that the watch

50
00:03:13,076 --> 00:03:15,796
Speaker 1: can talk to the world. If the antenna isn't just right,

51
00:03:16,036 --> 00:03:19,516
Speaker 1: the watch doesn't work. Yeah, So let's say hypothetically, you

52
00:03:19,676 --> 00:03:21,156
Speaker 1: at the end of the line, you have a pile

53
00:03:21,196 --> 00:03:24,916
Speaker 1: of units that the antennas don't work. The antenna's really challenging.

54
00:03:24,956 --> 00:03:27,996
Speaker 1: It's like near a lot of metal and antenna's a

55
00:03:28,036 --> 00:03:32,076
Speaker 1: metal don't don't mix, and so the antennas don't work.

56
00:03:32,276 --> 00:03:36,636
Speaker 1: What you have as an engineer is test station results

57
00:03:36,676 --> 00:03:39,476
Speaker 1: that say this signal is low. It could be an

58
00:03:39,516 --> 00:03:42,956
Speaker 1: eighty percent tissue, like eighty eight out of ten don't work.

59
00:03:43,276 --> 00:03:46,596
Speaker 1: It could be a you know, point one percent issue.

60
00:03:46,756 --> 00:03:48,676
Speaker 1: You only have one that doesn't work. It made a

61
00:03:48,716 --> 00:03:51,236
Speaker 1: thousand and nine hundred and ninety nine of them work,

62
00:03:51,396 --> 00:03:53,356
Speaker 1: But what's wrong with this one? Because if you make

63
00:03:53,396 --> 00:03:55,156
Speaker 1: a million, what's wrong with that one? If you make

64
00:03:55,196 --> 00:03:58,596
Speaker 1: a million, one in a thousand is a big problem, right, yes,

65
00:03:58,876 --> 00:04:04,516
Speaker 1: exactly exactly, And so in development you care about every failure. So, okay,

66
00:04:04,516 --> 00:04:07,556
Speaker 1: so you have this problem the antenna doesn't work whatever

67
00:04:07,836 --> 00:04:10,116
Speaker 1: often or once in a while, what do you do?

68
00:04:10,276 --> 00:04:13,796
Speaker 1: We would gather up those physical units, which is non

69
00:04:13,836 --> 00:04:16,116
Speaker 1: trivial to get all those units in one place in

70
00:04:16,116 --> 00:04:20,316
Speaker 1: this big factory, and then we would very carefully start

71
00:04:20,356 --> 00:04:22,996
Speaker 1: to rip them apart to try to get more data

72
00:04:23,036 --> 00:04:26,076
Speaker 1: about what's wrong with these units? Why don't they work? Now.

73
00:04:26,236 --> 00:04:30,636
Speaker 1: One of the things that's interesting about antenna specifically and Watch,

74
00:04:30,716 --> 00:04:33,276
Speaker 1: if you've ever seen a teardown of the watch, is

75
00:04:33,556 --> 00:04:36,556
Speaker 1: when you're taking it apart, you could very easily be

76
00:04:36,676 --> 00:04:40,156
Speaker 1: destroying the evidence of the thing you're trying to find. Oh,

77
00:04:40,196 --> 00:04:43,396
Speaker 1: like you just budge the antenna a tiny little millimeter

78
00:04:43,516 --> 00:04:46,996
Speaker 1: or whatever, and that matter, yes, and it matters. And

79
00:04:47,076 --> 00:04:49,956
Speaker 1: so you might try to take some X ray images

80
00:04:49,996 --> 00:04:53,876
Speaker 1: before you disassemble, so you have some idea literally there

81
00:04:53,956 --> 00:04:57,956
Speaker 1: you have an X ray camera, yes, which is like

82
00:04:57,956 --> 00:05:02,236
Speaker 1: a three dimensional X ray, And then you take them apart,

83
00:05:02,476 --> 00:05:05,116
Speaker 1: and you'd very carefully try to figure out what was

84
00:05:05,156 --> 00:05:08,596
Speaker 1: going wrong. And sometimes those things are obvious. You open

85
00:05:08,676 --> 00:05:11,036
Speaker 1: it up, it has no antenna in it, So that's

86
00:05:11,036 --> 00:05:14,836
Speaker 1: why the antenna doesn't where. Sometimes it's very obvious. Sometimes

87
00:05:14,836 --> 00:05:17,956
Speaker 1: it's not obvious, and like, as a skilled engineer, you're

88
00:05:17,996 --> 00:05:19,996
Speaker 1: looking at it, You're like, I have no idea, And

89
00:05:20,036 --> 00:05:23,076
Speaker 1: then you start looking at the process. You go, you walk,

90
00:05:23,276 --> 00:05:25,716
Speaker 1: you know, fifty meters up the line to look at

91
00:05:25,716 --> 00:05:28,396
Speaker 1: where they're putting the antenna's in, and you're watching that

92
00:05:28,436 --> 00:05:31,556
Speaker 1: process and trying to see is there something happening here?

93
00:05:31,676 --> 00:05:35,236
Speaker 1: Are there like are there screwdrivers swinging across the line

94
00:05:35,236 --> 00:05:36,996
Speaker 1: that are like going to hit the parts and knock

95
00:05:37,076 --> 00:05:39,196
Speaker 1: the antenna? I mean, I feel like there's sort of

96
00:05:39,276 --> 00:05:44,836
Speaker 1: two layers of problems here. Right. There's the fundamental problem of, oh,

97
00:05:44,916 --> 00:05:47,596
Speaker 1: we manufactured a million of these things and we know

98
00:05:47,796 --> 00:05:49,796
Speaker 1: ten thousand of them are going to break, right. That

99
00:05:49,956 --> 00:05:55,076
Speaker 1: is just a straight up problem. There is also the

100
00:05:55,316 --> 00:05:59,556
Speaker 1: like engineers like you have to go fly to China,

101
00:05:59,636 --> 00:06:01,876
Speaker 1: you know, stand on an assembly line and just hope

102
00:06:01,876 --> 00:06:05,116
Speaker 1: they catch things, hope they figure out where the problem is.

103
00:06:05,316 --> 00:06:07,676
Speaker 1: Not even they're looking for problems, they don't even know

104
00:06:07,676 --> 00:06:10,516
Speaker 1: their problems are yet, Like I would go to a

105
00:06:10,636 --> 00:06:13,596
Speaker 1: build and it would be a successful build if we

106
00:06:13,716 --> 00:06:16,756
Speaker 1: left the build with one hundred issues in a spreadsheet

107
00:06:16,836 --> 00:06:19,356
Speaker 1: of different things that we needed to fix. So that

108
00:06:19,476 --> 00:06:22,756
Speaker 1: was the success. Like, the more problems you find, the better.

109
00:06:22,916 --> 00:06:24,836
Speaker 1: I mean, it's a success because you found it because

110
00:06:24,836 --> 00:06:26,716
Speaker 1: you know the problems are there, and you're just worried

111
00:06:26,756 --> 00:06:28,556
Speaker 1: that you're not going to find. Yes, that's like your

112
00:06:28,636 --> 00:06:31,996
Speaker 1: real nightmare problem if you're like a design engineer, right,

113
00:06:32,036 --> 00:06:34,196
Speaker 1: the one that like you don't know is a problem

114
00:06:34,276 --> 00:06:37,036
Speaker 1: until you've shipped a million pairs of earbuds or something.

115
00:06:37,476 --> 00:06:40,076
Speaker 1: The technical term for that is an escape. So that's

116
00:06:40,116 --> 00:06:44,716
Speaker 1: a failure that has escaped the problem escaped the factory. Yes, yes,

117
00:06:44,796 --> 00:06:50,596
Speaker 1: and so escapes are what essentially then cause returns and

118
00:06:50,676 --> 00:06:53,396
Speaker 1: if we go back to that waste, yeah, which is

119
00:06:53,476 --> 00:06:55,836
Speaker 1: hugely costly. Right, it's bad for the reputation of the

120
00:06:55,876 --> 00:06:59,716
Speaker 1: manufacturer because it's like, oh, it's a crappy product, bad reviews,

121
00:06:59,796 --> 00:07:03,396
Speaker 1: not a reliable problem Amazon, bad reviews, and you've got

122
00:07:03,396 --> 00:07:05,476
Speaker 1: to like refund the money, send the person a new one.

123
00:07:05,516 --> 00:07:07,876
Speaker 1: So it's costly on many levels. Yes, yes, and so

124
00:07:08,036 --> 00:07:13,396
Speaker 1: returns should be avoided, and you avoid returns by reducing escapes,

125
00:07:13,796 --> 00:07:18,676
Speaker 1: and you reduce escapes not only by figuring out all

126
00:07:18,676 --> 00:07:21,076
Speaker 1: the things that you can possibly test for. And so

127
00:07:21,116 --> 00:07:23,596
Speaker 1: that's why there are testations on the line, but the

128
00:07:23,636 --> 00:07:26,796
Speaker 1: testation can't test for everything. Did you just live in

129
00:07:26,916 --> 00:07:34,396
Speaker 1: fear all the time of missing some terrible problem? I

130
00:07:34,396 --> 00:07:38,116
Speaker 1: don't even know how to answer that. I think that

131
00:07:38,316 --> 00:07:42,636
Speaker 1: a good engineer is a paranoid engineer. So for me,

132
00:07:42,756 --> 00:07:45,316
Speaker 1: as an engineer, I thought it was kind of silly

133
00:07:45,356 --> 00:07:47,676
Speaker 1: that there is so much luck involved. You just have

134
00:07:47,716 --> 00:07:49,876
Speaker 1: to get lucky and catch the problem before it goes

135
00:07:49,876 --> 00:07:53,196
Speaker 1: into production. Luck like that, yes, and so you're trying

136
00:07:53,196 --> 00:07:56,076
Speaker 1: to find you're relying on luck to find things that

137
00:07:56,156 --> 00:08:00,556
Speaker 1: you can improve in development. And then there's also like

138
00:08:00,596 --> 00:08:03,516
Speaker 1: the scramble and the heroism that happens when you do

139
00:08:03,636 --> 00:08:06,076
Speaker 1: have a problem that everybody learns about. Oh, we have

140
00:08:06,156 --> 00:08:09,836
Speaker 1: this problem. So then you have this fired drill where

141
00:08:09,836 --> 00:08:12,076
Speaker 1: everybody flies in tries to figure out what's going on.

142
00:08:12,116 --> 00:08:14,996
Speaker 1: There's tons of freneticism and activity, and this just seemed

143
00:08:15,036 --> 00:08:16,676
Speaker 1: like a waste, Like why is it that we're so

144
00:08:16,756 --> 00:08:20,396
Speaker 1: reactive versus proactive? Why is it that like I don't

145
00:08:20,436 --> 00:08:22,796
Speaker 1: have access to the data that I need as an

146
00:08:22,796 --> 00:08:25,676
Speaker 1: engineer to actually proactively solve these problems. I have to

147
00:08:25,756 --> 00:08:30,116
Speaker 1: go and like hope versus No. After the break, Annah

148
00:08:30,196 --> 00:08:34,196
Speaker 1: launches Instrumental, a company that's trying to help engineers solve

149
00:08:34,236 --> 00:08:38,316
Speaker 1: the problem of finding problems. Help him go from hope

150
00:08:38,876 --> 00:08:49,556
Speaker 1: to know. That's the end of the ads. Now we're

151
00:08:49,556 --> 00:08:51,916
Speaker 1: going back to the show, So just a reset here.

152
00:08:52,276 --> 00:08:55,516
Speaker 1: Anna has landed a job at Apple, the company where

153
00:08:55,556 --> 00:09:00,156
Speaker 1: engineers dream of working, this amazing technology company, and yet

154
00:09:00,196 --> 00:09:02,796
Speaker 1: she decides to leave. So I asked her in the end,

155
00:09:02,916 --> 00:09:05,356
Speaker 1: what made you quit your job and go start this

156
00:09:05,396 --> 00:09:10,156
Speaker 1: new company, Instrumental. I can share some thing that that

157
00:09:10,316 --> 00:09:12,436
Speaker 1: you can use, but it's maybe it might be a

158
00:09:12,436 --> 00:09:14,636
Speaker 1: little sideways to what you're talking about, but it's the

159
00:09:14,676 --> 00:09:18,436
Speaker 1: true reason I started the company. So about a year

160
00:09:18,436 --> 00:09:21,436
Speaker 1: before Apple Watch shipped and the world knew that it

161
00:09:21,516 --> 00:09:24,476
Speaker 1: even existed, I actually had a personal tragedy happened in

162
00:09:24,476 --> 00:09:27,676
Speaker 1: my life. My husband was killed by a drunk driver.

163
00:09:28,236 --> 00:09:31,356
Speaker 1: Oh my god, and I was twenty seven, he was

164
00:09:31,396 --> 00:09:39,636
Speaker 1: twenty seven. And I realized that life is precious and limited,

165
00:09:40,556 --> 00:09:42,276
Speaker 1: and you don't know when your last day is going

166
00:09:42,316 --> 00:09:44,676
Speaker 1: to be, and so you want to be proud of

167
00:09:44,716 --> 00:09:47,036
Speaker 1: the days that you have and how you spent them

168
00:09:47,196 --> 00:09:49,636
Speaker 1: and the impact you could have on others, and also

169
00:09:49,716 --> 00:09:54,836
Speaker 1: reevaluate the justification for my existence on this planet. And

170
00:09:55,356 --> 00:09:58,676
Speaker 1: I decided that Apple is great. I'm glad I was there,

171
00:09:59,036 --> 00:10:01,596
Speaker 1: but it can't be my life's work for me. I

172
00:10:01,716 --> 00:10:05,876
Speaker 1: felt like I needed to do something bigger. And the

173
00:10:06,036 --> 00:10:08,916
Speaker 1: big problem that we all need to solve is how

174
00:10:09,156 --> 00:10:12,956
Speaker 1: we build stuff is so wasteful. That is a more

175
00:10:13,076 --> 00:10:17,036
Speaker 1: meaningful direction. And there is a big problem that I

176
00:10:17,076 --> 00:10:19,916
Speaker 1: know that I have some intuition about how to go solve,

177
00:10:20,156 --> 00:10:23,116
Speaker 1: And so I took what I understood as like, wow,

178
00:10:23,116 --> 00:10:24,876
Speaker 1: it's like really hard to find and fix these issues.

179
00:10:24,876 --> 00:10:28,116
Speaker 1: We're wasting so much money, so much physical stuff. We

180
00:10:28,196 --> 00:10:32,076
Speaker 1: are pumping so much chemicals into our rivers, we are

181
00:10:32,196 --> 00:10:36,716
Speaker 1: burning so much energy. We are wasting human lifetimes of

182
00:10:36,796 --> 00:10:44,596
Speaker 1: time doing things that don't matter. It's just wasted, and

183
00:10:45,236 --> 00:10:47,276
Speaker 1: we need to figure that out. We need to change

184
00:10:47,276 --> 00:10:50,156
Speaker 1: how we build, We need to change how we think

185
00:10:50,196 --> 00:10:53,076
Speaker 1: about building products. So this is like a little maybe

186
00:10:53,396 --> 00:10:56,556
Speaker 1: sappy for your audience, but this is the true reason

187
00:10:57,476 --> 00:11:02,516
Speaker 1: that like really made the change of thinking about away

188
00:11:02,596 --> 00:11:04,836
Speaker 1: from thinking about, oh, my job as an engineer, I

189
00:11:04,996 --> 00:11:08,876
Speaker 1: make stuff too. Oh, my job as a human is

190
00:11:08,876 --> 00:11:11,956
Speaker 1: to figure out how to do the things we need

191
00:11:12,036 --> 00:11:15,396
Speaker 1: to do better. So Anna decides she's going to try

192
00:11:15,436 --> 00:11:20,996
Speaker 1: to make electronics manufacturing better, more efficient, less wasteful. And

193
00:11:21,076 --> 00:11:23,356
Speaker 1: to do this, she and an engineer she worked with

194
00:11:23,396 --> 00:11:26,596
Speaker 1: at Apple decide to start a company. They call it Instrumental,

195
00:11:27,196 --> 00:11:30,956
Speaker 1: And when it started, they had one big idea. Instead

196
00:11:30,996 --> 00:11:33,996
Speaker 1: of forcing engineers like Anna to frantically run up and

197
00:11:33,996 --> 00:11:37,676
Speaker 1: down the assembly line hoping they'd spotted every potential problem,

198
00:11:38,196 --> 00:11:41,156
Speaker 1: they would put dozens of cameras up and down the

199
00:11:41,196 --> 00:11:46,476
Speaker 1: line to capture the entire assembly process for every single device.

200
00:11:46,876 --> 00:11:50,436
Speaker 1: It's the idea that there is value in this data

201
00:11:50,676 --> 00:11:54,956
Speaker 1: that today does not get collected, and so the actual

202
00:11:54,996 --> 00:11:58,476
Speaker 1: core pieces, Let's go get all the data. That's the

203
00:11:58,916 --> 00:12:01,836
Speaker 1: that's the core innovation here. Let's go get all the data.

204
00:12:02,036 --> 00:12:04,836
Speaker 1: The data in this case is visual data. What does

205
00:12:04,876 --> 00:12:07,996
Speaker 1: everything look like as this phone or this watch or

206
00:12:08,036 --> 00:12:10,356
Speaker 1: whatever is getting put together? Is that right? That was

207
00:12:10,396 --> 00:12:13,076
Speaker 1: the first data set that we went after was the

208
00:12:13,116 --> 00:12:17,196
Speaker 1: image data set, because there isn't one. So we're creating

209
00:12:17,436 --> 00:12:20,156
Speaker 1: a data set that didn't exist before. And as you know,

210
00:12:20,276 --> 00:12:22,516
Speaker 1: as they say, a picture is worth a thousand words.

211
00:12:23,236 --> 00:12:24,956
Speaker 1: When you take a picture, you don't need to know

212
00:12:24,996 --> 00:12:27,476
Speaker 1: in advance what you're going to be looking for in

213
00:12:27,516 --> 00:12:31,196
Speaker 1: the picture. It's a very high resolution data set, and

214
00:12:31,236 --> 00:12:34,196
Speaker 1: then you can come back with twenty twenty hindsight and

215
00:12:34,236 --> 00:12:36,796
Speaker 1: look at the picture and be like, oh, like, here's

216
00:12:36,796 --> 00:12:39,316
Speaker 1: a problem in that picture. And so that was the

217
00:12:39,356 --> 00:12:41,956
Speaker 1: initial concept of like, we need to find where these

218
00:12:41,956 --> 00:12:43,996
Speaker 1: issues are, which means we need to see them. We

219
00:12:44,076 --> 00:12:46,116
Speaker 1: work in the physical world, so we need to see them.

220
00:12:46,356 --> 00:12:48,556
Speaker 1: Remember I was talking about the antennas, where you had

221
00:12:48,596 --> 00:12:51,356
Speaker 1: to carefully take them apart, and as you took them apart,

222
00:12:51,436 --> 00:12:54,956
Speaker 1: you'd sometimes damage the evidence you're looking for. Well, if

223
00:12:54,956 --> 00:12:58,036
Speaker 1: you have images of them as they went together, then

224
00:12:58,196 --> 00:13:00,236
Speaker 1: you don't maybe have to take them apart, or you

225
00:13:00,316 --> 00:13:03,876
Speaker 1: maybe have additional data about what they looked like as

226
00:13:03,876 --> 00:13:05,796
Speaker 1: they were going together. But you don't know which ones

227
00:13:05,836 --> 00:13:08,316
Speaker 1: are going to be failures in advance, so you take

228
00:13:08,316 --> 00:13:10,596
Speaker 1: pictures of all of them and then you have the

229
00:13:10,636 --> 00:13:13,076
Speaker 1: ones after the fact that ended up being failures. You

230
00:13:13,116 --> 00:13:15,796
Speaker 1: have those images. So that was like the first idea.

231
00:13:15,916 --> 00:13:18,036
Speaker 1: So the basic ideas. You take pictures of the whole thing,

232
00:13:18,116 --> 00:13:20,756
Speaker 1: and you identify, you know, whatever, the serial number on

233
00:13:20,836 --> 00:13:24,116
Speaker 1: each say, watch each device that's going down the line,

234
00:13:24,276 --> 00:13:26,796
Speaker 1: and then if there's one device that doesn't work, you say,

235
00:13:26,796 --> 00:13:28,196
Speaker 1: show me all the pictures of that one, and then

236
00:13:28,196 --> 00:13:29,916
Speaker 1: you say, how is that one different than all the

237
00:13:29,916 --> 00:13:32,756
Speaker 1: ones that do work in the pictures? Yes, yes, But

238
00:13:33,036 --> 00:13:35,596
Speaker 1: it's not just about figuring out what's wrong once you

239
00:13:35,636 --> 00:13:38,156
Speaker 1: find a problem, right is It also about getting better

240
00:13:38,196 --> 00:13:41,236
Speaker 1: at finding problems in the first place. So discovery, we're

241
00:13:41,236 --> 00:13:43,956
Speaker 1: able to solve the discovery problem for our customer. We

242
00:13:44,076 --> 00:13:50,156
Speaker 1: built algorithms that look for anomalies, and so what's an anomaly.

243
00:13:50,716 --> 00:13:53,796
Speaker 1: An anomaly is something that's different from the other ones.

244
00:13:54,036 --> 00:13:56,916
Speaker 1: So if you build a population of units, they should

245
00:13:56,916 --> 00:13:59,636
Speaker 1: all look the same, Okay, a bunch of devices, right,

246
00:13:59,876 --> 00:14:03,796
Speaker 1: The ones that look different are probably interesting, and so

247
00:14:03,836 --> 00:14:07,436
Speaker 1: we build algorithms to find the ones that look different.

248
00:14:07,836 --> 00:14:11,196
Speaker 1: And that was kind of the first offering that we provided,

249
00:14:11,396 --> 00:14:14,156
Speaker 1: And so our customers were able to use that combination

250
00:14:14,716 --> 00:14:19,636
Speaker 1: of the photo record with the algorithms that would highlight

251
00:14:19,676 --> 00:14:23,156
Speaker 1: automatically the things that were different and interesting, and they

252
00:14:23,156 --> 00:14:26,436
Speaker 1: were able to use that to find issues. Basically, one

253
00:14:26,476 --> 00:14:29,236
Speaker 1: of these things is not like the other, yes, exactly,

254
00:14:29,356 --> 00:14:31,996
Speaker 1: So like, oh, this one's bent a little bit, this

255
00:14:32,036 --> 00:14:34,396
Speaker 1: one's the parts supposed to be black and it's white.

256
00:14:34,476 --> 00:14:37,916
Speaker 1: Is that a problem? Maybe not? Maybe we highlight those

257
00:14:37,956 --> 00:14:41,716
Speaker 1: things and then the next piece is okay, well what's

258
00:14:41,756 --> 00:14:45,436
Speaker 1: causing like problems? Okay, so you've built a system that

259
00:14:45,556 --> 00:14:49,636
Speaker 1: is better at identifying for finding problems. So now even

260
00:14:49,676 --> 00:14:53,436
Speaker 1: if a device pass all the tests, your system will say, well,

261
00:14:53,436 --> 00:14:55,956
Speaker 1: you want to check out this one missing screws because

262
00:14:55,956 --> 00:14:59,356
Speaker 1: it just looks a little different missing screws. Okay, so yeah,

263
00:14:59,476 --> 00:15:02,396
Speaker 1: like that missing screws, there's no screw tester. You're right,

264
00:15:02,436 --> 00:15:05,996
Speaker 1: like missing screws. Devices will actually miss screws. I'm so

265
00:15:06,116 --> 00:15:07,916
Speaker 1: naive that I'm like, well, surely a device wouldn't be

266
00:15:07,996 --> 00:15:10,996
Speaker 1: missing a screw, but one hundred percent, it's like one

267
00:15:11,036 --> 00:15:13,796
Speaker 1: of the top ten defects in production is missing or

268
00:15:13,996 --> 00:15:17,396
Speaker 1: extra parts, extra screw just to make up for the

269
00:15:17,396 --> 00:15:19,396
Speaker 1: other one that's missing a screw. On average, they have

270
00:15:19,436 --> 00:15:22,436
Speaker 1: the right number of screws. Yes, on average they have

271
00:15:22,516 --> 00:15:25,716
Speaker 1: the right number of scares. So we need to be

272
00:15:25,756 --> 00:15:28,156
Speaker 1: able to find issues. We need to be able to

273
00:15:28,196 --> 00:15:31,716
Speaker 1: help engineers fix them, and then we need to help

274
00:15:31,756 --> 00:15:33,756
Speaker 1: them make sure they don't come back. Those are kind

275
00:15:33,756 --> 00:15:36,516
Speaker 1: of the three things that we have to do. So okay,

276
00:15:36,596 --> 00:15:39,196
Speaker 1: those are the three key things Anna's company does to

277
00:15:39,236 --> 00:15:42,596
Speaker 1: solve this big problem and to tile those pieces together.

278
00:15:42,916 --> 00:15:45,436
Speaker 1: She gave me a case study, a case study from

279
00:15:45,436 --> 00:15:48,396
Speaker 1: a company that was trying to build a webcam and

280
00:15:48,476 --> 00:15:51,596
Speaker 1: the company was having trouble with the antenna. The antenna, which,

281
00:15:51,756 --> 00:15:54,516
Speaker 1: as Anna told me about earlier in the interview, is

282
00:15:54,556 --> 00:15:57,676
Speaker 1: this crucial part that is surprisingly hard to get right.

283
00:15:57,876 --> 00:16:01,036
Speaker 1: They always fail. There are very sensitive parts, and all

284
00:16:01,036 --> 00:16:04,036
Speaker 1: of our connected devices really rely on their antenna performance.

285
00:16:04,036 --> 00:16:06,116
Speaker 1: So you don't think about it, but you know it's

286
00:16:06,156 --> 00:16:08,716
Speaker 1: an important part. And so we had a customer who

287
00:16:08,756 --> 00:16:13,116
Speaker 1: had failing antennas and they were able to take that

288
00:16:13,236 --> 00:16:15,756
Speaker 1: group of that population of units and they were able

289
00:16:15,796 --> 00:16:18,596
Speaker 1: to actively, like root cause it to three different things.

290
00:16:18,636 --> 00:16:21,236
Speaker 1: And so this is actually the next piece of the puzzle,

291
00:16:21,516 --> 00:16:24,476
Speaker 1: which is we need to find root cause. And so

292
00:16:24,916 --> 00:16:27,436
Speaker 1: in this case, the customer was able to see a

293
00:16:27,476 --> 00:16:29,956
Speaker 1: couple of different things. The first thing they found is

294
00:16:29,996 --> 00:16:32,516
Speaker 1: a group of those failures was actually had to do

295
00:16:32,676 --> 00:16:36,276
Speaker 1: with the alignment of a connector that was connecting the antenna.

296
00:16:36,356 --> 00:16:38,556
Speaker 1: So the connector, if it was like shifted a little

297
00:16:38,596 --> 00:16:40,796
Speaker 1: bit this way, it was good, and it was shifted

298
00:16:40,836 --> 00:16:43,116
Speaker 1: a little bit the other way, it was bad. And

299
00:16:43,156 --> 00:16:45,836
Speaker 1: this was very subtle. This is not something they had

300
00:16:45,916 --> 00:16:50,916
Speaker 1: like a specification. This is a discovery. Okay, so the

301
00:16:51,036 --> 00:16:53,916
Speaker 1: angle mattered and they didn't know that in advance. Okay,

302
00:16:53,996 --> 00:16:57,556
Speaker 1: Now this is also maybe too much in the weeds.

303
00:16:57,636 --> 00:17:00,236
Speaker 1: But there are a lot of reasons, and antenna can

304
00:17:00,276 --> 00:17:02,636
Speaker 1: fail in a lot of different ways. So it's not

305
00:17:02,716 --> 00:17:05,436
Speaker 1: just like, oh, if they're failing, they're all the same

306
00:17:05,836 --> 00:17:08,916
Speaker 1: root costs. Actually, there could be a group of different

307
00:17:09,236 --> 00:17:12,196
Speaker 1: causes that cause failures. So you have two different phones

308
00:17:12,396 --> 00:17:15,276
Speaker 1: from the same build, both of which have the antenna

309
00:17:15,316 --> 00:17:17,476
Speaker 1: that doesn't work. It might be a different reason for

310
00:17:17,516 --> 00:17:19,756
Speaker 1: each phone that the antenna doesn't work, which makes it

311
00:17:19,756 --> 00:17:21,516
Speaker 1: harder to solve. So you solve one problem, but then

312
00:17:21,556 --> 00:17:22,916
Speaker 1: some of the phones still don't work. Is that what

313
00:17:22,916 --> 00:17:26,276
Speaker 1: you're saying, Yes, yes, that's very common. They also found

314
00:17:26,356 --> 00:17:29,796
Speaker 1: that the software version that they were using on the

315
00:17:29,836 --> 00:17:33,396
Speaker 1: tester was also correlated to the failures, so most of

316
00:17:33,436 --> 00:17:36,196
Speaker 1: the failures were from one particular tester, So in fact,

317
00:17:36,236 --> 00:17:39,916
Speaker 1: it wasn't a bad antenna. It was a test for

318
00:17:39,996 --> 00:17:41,836
Speaker 1: some of those units, not all of them. Some of

319
00:17:41,836 --> 00:17:44,116
Speaker 1: them still had the connector coming out at a wrong angle.

320
00:17:44,236 --> 00:17:47,636
Speaker 1: Yeah yeah. And then the last one on the same

321
00:17:47,716 --> 00:17:50,316
Speaker 1: group of failures, in the same group of failures, there

322
00:17:50,396 --> 00:17:54,756
Speaker 1: was another group of units where what the image based

323
00:17:54,756 --> 00:17:57,716
Speaker 1: algorithms found as a high correlation to the color of

324
00:17:57,796 --> 00:18:02,356
Speaker 1: the circuit board. And so color doesn't affect performance, but

325
00:18:02,876 --> 00:18:05,756
Speaker 1: it was a different vendor. One circuit board was like

326
00:18:05,796 --> 00:18:09,316
Speaker 1: a slightly different shade of a color than another, and

327
00:18:09,436 --> 00:18:11,796
Speaker 1: so that means like the vendor of that circuit board

328
00:18:12,996 --> 00:18:15,476
Speaker 1: was for one of the vendors that made that circuit

329
00:18:15,476 --> 00:18:17,516
Speaker 1: board was more likely to have failing in it. So

330
00:18:17,596 --> 00:18:20,236
Speaker 1: that gives engineers then three different things. They can go

331
00:18:20,356 --> 00:18:22,556
Speaker 1: chase down. They can go chase down a process is

332
00:18:22,636 --> 00:18:24,876
Speaker 1: year around the angle of the connector. They can go

333
00:18:24,996 --> 00:18:28,476
Speaker 1: chase down like the drifting test station and work on

334
00:18:29,436 --> 00:18:32,476
Speaker 1: calibrating the testations. And then they can go and investigate

335
00:18:32,636 --> 00:18:35,636
Speaker 1: if there's something meaningfully different between these two vendors of

336
00:18:35,836 --> 00:18:38,316
Speaker 1: circuit boards. So they're able to continue building, but they

337
00:18:38,316 --> 00:18:41,236
Speaker 1: have now three specific things they can go do, whereas

338
00:18:41,276 --> 00:18:43,876
Speaker 1: without instrumental they would kind of be guessing. So, like,

339
00:18:44,316 --> 00:18:46,436
Speaker 1: what's the sort of frontier for you? What's the thing

340
00:18:46,476 --> 00:18:48,916
Speaker 1: you're trying to do that you haven't quite figured out yet, Like,

341
00:18:48,916 --> 00:18:51,956
Speaker 1: what's the next problem to solve? Yeah, I mean so

342
00:18:52,036 --> 00:18:54,676
Speaker 1: it goes back to the reason for being for the company,

343
00:18:54,996 --> 00:18:57,876
Speaker 1: the reason for being being that twenty cents of every

344
00:18:57,876 --> 00:19:01,156
Speaker 1: dollar that's wasted in manufacturing, we haven't solved that problem yet.

345
00:19:01,316 --> 00:19:04,516
Speaker 1: What does that require? We need to change how we

346
00:19:04,556 --> 00:19:07,036
Speaker 1: think about how we build things, how we design things,

347
00:19:07,116 --> 00:19:11,356
Speaker 1: the process of iterating through that development process, the process

348
00:19:11,396 --> 00:19:14,156
Speaker 1: of what happens when it goes into production, what happens

349
00:19:14,196 --> 00:19:16,316
Speaker 1: when you return units? And what do we do with

350
00:19:16,356 --> 00:19:20,796
Speaker 1: that information from your return as a consumer. And today

351
00:19:20,836 --> 00:19:23,836
Speaker 1: we're really not doing that as an industry. Not gathering

352
00:19:23,996 --> 00:19:27,276
Speaker 1: data in a meaningful way is that we're not gathering data,

353
00:19:27,436 --> 00:19:31,356
Speaker 1: we're not thinking proactively. As an engineer. If I'm designing something,

354
00:19:31,396 --> 00:19:33,996
Speaker 1: I'm thinking about, you know, the next build and production,

355
00:19:34,036 --> 00:19:38,276
Speaker 1: but I'm not I'm not necessarily thinking about, like how

356
00:19:38,316 --> 00:19:40,996
Speaker 1: can I develop this product's that I can capture the

357
00:19:41,036 --> 00:19:44,316
Speaker 1: most data from it the fastest, the earlier in the

358
00:19:44,356 --> 00:19:47,636
Speaker 1: process versus later in the product. Like everybody knows their

359
00:19:47,716 --> 00:19:50,156
Speaker 1: data is valuable and there's stuff in there that could

360
00:19:50,156 --> 00:19:52,876
Speaker 1: be valuable, but they don't know how to value it

361
00:19:53,116 --> 00:19:56,236
Speaker 1: because the problems haven't happened yet. When they're on fire

362
00:19:56,316 --> 00:20:00,396
Speaker 1: quote unquote and they are spending a million dollars a

363
00:20:00,436 --> 00:20:04,156
Speaker 1: month on returns, they know exactly what the value is

364
00:20:04,196 --> 00:20:07,356
Speaker 1: in the data that they don't have, Yes, but when

365
00:20:07,396 --> 00:20:10,116
Speaker 1: they haven't started the program yet and they don't know

366
00:20:10,156 --> 00:20:12,836
Speaker 1: what fires they're going to have, they don't know how

367
00:20:12,836 --> 00:20:16,356
Speaker 1: to value having the data. And like preventing those problems

368
00:20:16,356 --> 00:20:18,516
Speaker 1: in the first place, the vision is that that data

369
00:20:18,596 --> 00:20:21,956
Speaker 1: is enough and then if you figured out how to

370
00:20:21,996 --> 00:20:24,476
Speaker 1: harvest that data, you can actually build lines that improve

371
00:20:24,516 --> 00:20:27,756
Speaker 1: themselves and then you eliminate the waste. So this new

372
00:20:27,796 --> 00:20:30,196
Speaker 1: different way of doing things, like can you just tell

373
00:20:30,196 --> 00:20:32,876
Speaker 1: me sort of specifically, like a few details, what would

374
00:20:32,876 --> 00:20:35,036
Speaker 1: it be like, what would it look like, what would

375
00:20:35,076 --> 00:20:38,716
Speaker 1: be happening in this world? Everything gets built cheaper, everything

376
00:20:38,756 --> 00:20:42,996
Speaker 1: gets built with less waste. And when that happens, we

377
00:20:43,076 --> 00:20:45,716
Speaker 1: have maybe we have cheaper products, we have maybe more

378
00:20:45,716 --> 00:20:49,716
Speaker 1: profitable companies, we have less waste going into the world,

379
00:20:49,956 --> 00:20:53,756
Speaker 1: less physical waste, chemicals and rivers, less energy use, less

380
00:20:53,796 --> 00:20:58,636
Speaker 1: human lifetimes wasted, And it's just thinking completely differently about

381
00:20:58,756 --> 00:21:02,916
Speaker 1: like that manufacturing actually is a machine itself, and that

382
00:21:02,956 --> 00:21:05,356
Speaker 1: we need to optimize that whole process as a machine

383
00:21:05,356 --> 00:21:08,636
Speaker 1: itself versus a means to justify the ends of like

384
00:21:08,836 --> 00:21:10,836
Speaker 1: we need we need to get products out the other

385
00:21:10,916 --> 00:21:14,236
Speaker 1: end of the line. In a minute, the Lightning Round,

386
00:21:15,116 --> 00:21:18,316
Speaker 1: Anna tells us what mechanical engineers know about the world

387
00:21:18,876 --> 00:21:30,356
Speaker 1: and the inefficiencies out in the wild that grind her tears. Now,

388
00:21:30,436 --> 00:21:34,116
Speaker 1: let's get back to what's your problem? Great, let's do

389
00:21:34,156 --> 00:21:37,996
Speaker 1: the Lightning Round. Are you ready? What's one piece of

390
00:21:38,036 --> 00:21:41,276
Speaker 1: advice you'd give to someone trying to solve a hard problem.

391
00:21:41,356 --> 00:21:45,796
Speaker 1: I love that question. I love it. I had a

392
00:21:45,836 --> 00:21:47,716
Speaker 1: science teacher in high school taught me how to solve

393
00:21:47,756 --> 00:21:50,236
Speaker 1: hard problems because I was working on science research. I'm

394
00:21:50,276 --> 00:21:53,916
Speaker 1: a science fair kid, and he taught me to break

395
00:21:53,956 --> 00:21:56,996
Speaker 1: those problems down into manageable pieces. So like, if you

396
00:21:56,996 --> 00:22:00,836
Speaker 1: can take smaller steps and understanding a problem and the

397
00:22:00,916 --> 00:22:04,476
Speaker 1: small steps you can do, then it becomes easier to solve.

398
00:22:04,956 --> 00:22:08,276
Speaker 1: So as somebody whose job is sort of to find

399
00:22:08,516 --> 00:22:12,116
Speaker 1: errors and inefficiencies in the world. Is there some some

400
00:22:12,276 --> 00:22:15,076
Speaker 1: domain something you encounter in your daily life that you

401
00:22:15,196 --> 00:22:19,956
Speaker 1: just you really want to optimize inefficiencies in things that

402
00:22:20,036 --> 00:22:25,276
Speaker 1: cause lines really irk me, like at the airport or

403
00:22:25,276 --> 00:22:27,956
Speaker 1: in like a restaurant, like anything that feels like a

404
00:22:28,036 --> 00:22:31,796
Speaker 1: non optimal like kind of scheduling and line that causes

405
00:22:31,796 --> 00:22:34,276
Speaker 1: a line of humans. That is something that I that

406
00:22:34,356 --> 00:22:37,316
Speaker 1: I noticed, and I just like really grinds my gears.

407
00:22:37,796 --> 00:22:40,716
Speaker 1: Of all your patents, which one's your favorite, I'm gonna

408
00:22:40,796 --> 00:22:44,196
Speaker 1: go with my first patent more because of what it means.

409
00:22:44,756 --> 00:22:48,476
Speaker 1: So I share my first patent with actually who is

410
00:22:48,516 --> 00:22:52,756
Speaker 1: now my co founder, Sam Weiss, and we met in

411
00:22:52,916 --> 00:22:55,276
Speaker 1: two thousand and nine at Apple, and that's actually the

412
00:22:55,316 --> 00:22:57,876
Speaker 1: summer that we invented our mutual. It was our mutual

413
00:22:57,916 --> 00:23:02,236
Speaker 1: first pattern, and we had to build essentially a switch

414
00:23:02,396 --> 00:23:04,636
Speaker 1: very similar to you know on the side of an

415
00:23:04,636 --> 00:23:07,436
Speaker 1: iPhone there's like a little ringer switch that you can

416
00:23:08,236 --> 00:23:10,556
Speaker 1: you can kind of down. So we had to build

417
00:23:10,556 --> 00:23:14,636
Speaker 1: a switch like that that had a very small ford

418
00:23:14,676 --> 00:23:16,756
Speaker 1: back there, like it needed to fit in a weird shape.

419
00:23:17,076 --> 00:23:19,556
Speaker 1: And so that was our first pattern. That pattern will

420
00:23:19,596 --> 00:23:21,476
Speaker 1: never see the light of day. There will never be

421
00:23:21,516 --> 00:23:24,276
Speaker 1: a switch that has made in that design, but it

422
00:23:24,356 --> 00:23:29,476
Speaker 1: was cool. That's awesome. What do mechanical engineers know about

423
00:23:29,516 --> 00:23:32,276
Speaker 1: the world that nobody else really gets? That everything is

424
00:23:32,316 --> 00:23:35,956
Speaker 1: imperfect and different from every other thing. Is there a

425
00:23:35,956 --> 00:23:39,356
Speaker 1: particular piece in say my iPhone that is the one

426
00:23:39,436 --> 00:23:42,796
Speaker 1: that is like has caused the most manufacturing problems, the

427
00:23:42,796 --> 00:23:44,756
Speaker 1: most problems on the assembly line? Is there some piece

428
00:23:44,796 --> 00:23:47,116
Speaker 1: that's like, oh, that piece is the one. I never

429
00:23:47,156 --> 00:23:49,676
Speaker 1: worked on phones, so I don't know specifically, but in

430
00:23:49,716 --> 00:23:57,116
Speaker 1: general in products, the antenna, anna, anything with glue, and

431
00:23:57,836 --> 00:23:59,876
Speaker 1: anything having to do with water, So those are the

432
00:23:59,916 --> 00:24:02,516
Speaker 1: three things. It's like. It's like the antenna's heart displays

433
00:24:02,516 --> 00:24:06,756
Speaker 1: are hard too, but like, so basically everything. Is there

434
00:24:06,836 --> 00:24:10,476
Speaker 1: some trick that mechanical engineers use when something isn't working,

435
00:24:10,516 --> 00:24:13,476
Speaker 1: like a remote control or whatever. I always think about it,

436
00:24:13,556 --> 00:24:15,956
Speaker 1: is it a hardware problem or a software problem. If

437
00:24:15,956 --> 00:24:19,676
Speaker 1: it's a software problem, reboot it. If it's a hardware problem,

438
00:24:19,716 --> 00:24:21,876
Speaker 1: I don't know, check the batteries. I make sure it's

439
00:24:21,916 --> 00:24:29,716
Speaker 1: plugged in. Anna Katrina shad Letsky is the founder and

440
00:24:29,876 --> 00:24:34,196
Speaker 1: CEO of Instrumental. I have a request for you this week,

441
00:24:34,356 --> 00:24:37,716
Speaker 1: and it is this. Please let me know who you

442
00:24:37,756 --> 00:24:40,396
Speaker 1: want to hear on this show. You can email me

443
00:24:40,556 --> 00:24:44,956
Speaker 1: at problem at Pushkin dot fm. That's problem at Pushkin

444
00:24:45,036 --> 00:24:48,596
Speaker 1: dot Fm. Or you can tweet at me at Jacob Goldstein.

445
00:24:49,756 --> 00:24:53,796
Speaker 1: Today's show was edited by Robert Smith, produced by Edith Russolo,

446
00:24:53,956 --> 00:24:57,636
Speaker 1: and engineered by Amanda k Wong. I'm Jacob Goldstein, and

447
00:24:57,676 --> 00:25:00,036
Speaker 1: I'll be back next week with another episode of What's

448
00:25:00,036 --> 00:25:04,476
Speaker 1: Your Problem