UP037: Preteckt // real-time vehicle prognostics using machine learning

In All Episodes, SaaS, upside by jayclouseLeave a Comment

view episode transcript

Ken Sills 0:00
You got on the phone with me is I’m going to tell you this is going to be a very aggressive call. I’ve had so many people blowing smoke up my ass or less us about this ml stuff, solving it. And none of it works like, Oh, yeah, you’re right, it doesn’t. So, you know, here’s, here’s where we’re different. And, you know, number one, we take boatloads of data. No one else has taken that kind of data and honestly, garbage and equals garbage out. That’s a fundamental tenet of science, right? So you have to take great data. And so we entered this as a data science program, right? Like to me, I’m a data scientist. First thing was how do I get great data. And since we couldn’t find anybody who had a piece of hardware that would give us great data. So we’re making that piece of hardware.

Jay Clouse 0:37
The startup investment landscape is changing and world class companies are being built outside of Silicon Valley. We find them talk with them and discuss the upside of investing in them.

Welcome to upside.

Eric Hornung 1:04
Hello. Hello. Hello and welcome to the upside podcast first podcast, finding upside outside of Silicon Valley. I’m Eric Warren, and I’m accompanied by my co host, Mr. internal monologue himself. Jay Klaus. Jay, how’s it going, man? You reminder me of Austin Powers sometimes.

Jay Clouse 1:22
What do you. What do you know about my internal monologue?

Eric Hornung 1:24
I know that you don’t have one. I don’t think you have one.

Jay Clouse 1:26
Oh, I have one. You don’t think I have an internal monologue?

Eric Hornung 1:29
I don’t know, man. We were walking around CES and you. You said some things out loud that I would have said in my head.

Jay Clouse 1:36
Oh, you’re saying that I should use more of an internal monologue as opposed to speaking my thoughts out loud.

Eric Hornung 1:41
Not all the time. But sometimes, yeah.

Jay Clouse 1:43

That, that does happen by accident. Sometimes. Sometimes. I think I’m thinking and I’m actually speaking and I definitely speak out loud in public places. Just something that I’m thinking and sometimes it’s real weird.

Eric Hornung 1:54
What’s the weirdest thing you’ve ever said out loud that you thought you were thinking?

Jay Clouse 1:57
I don’t know as it be hard for me to tell. But sometimes I rehearse conversations before they happen, so people will just randomly hear me speaking to somebody and there’s nobody there.

Eric Hornung 2:06
Do you do that? Like when you’re walking to a coffee shop? When do you do that?

Jay Clouse 2:09
When I’m walking to a coffee shop? When I’m having conflict with somebody that I haven’t resolved, it could be anywhere really

Eric Hornung 2:15
interesting. See, that seems more productive to me, then my initial jab. I think that rehearsing your conversations actually makes a lot of sense.

Jay Clouse 2:24
Yeah, yeah. No, it’s good. Sometimes it’s useful sometimes. And then a lot of times just really weird and off putting. So yeah, thanks for calling me out is definitely not appropriate at CES when people are showcasing their life’s work and I accidentally say out loud “don’t like that!”

Eric Hornung 2:42
these people are like they have their whole family there they’re building this company from scratch like “nah”

Jay Clouse 2:49
You’ve probably been working on this technology for years and submitted thousands of dollars worth of patents and it’s just not at all interesting to me.

Eric Hornung 2:55
Well, you know, at least you’re honest, that’s that’s big. So let’s transition into our guests today.

Jay Clouse 3:00
Right into it, I like it. Let’s, let’s keep this one tight. Today we’re talking to Ken sills, the CEO of Preteckt, Preteckt uses machine learning to predict vehicle service needs before they get serious trucks on their network. learn from each other so that their customers can predict breakdowns weeks or months before they cause downtime. They’re connected hardware gives real time unparalleled insight into vehicle maintenance needs. It’s like driving with a master mechanic in your engine compartment. pretext is based in Memphis, Tennessee, founded in 2015 and referred to us by our friend Katie Milligan at Start Co. Ken’s got an interesting background that I’m excited to talk about, because he was formerly the CTO of Preteckt is now the CEO. He was the one that developed the initial prototype of the hardware and software solution in 2015 and has built out their team of engineers, scientists, business developers and sales professionals. So I’m interested to hear that transition for can from CTO to CEO, both why it happened and the practical side of how it works. Did you get some reminiscence of super dispatch As you’re researching here, Eric?

Eric Hornung 4:05
I did I I felt that it’s an industry that we’ve already kind of learned a little bit about, which is nice, because a lot of times on upside will be coming into an industry fresh. And that comes with a huge learning curve. But now we kind of have a little bit of a background. So I think that our questions may have a little bit more depth to them. Then unusual. Yeah, like you learning about some of these hard technology companies that have both hardware and software because in some ways that’s building two products almost right. So like to hear more about that they’ve raised from what I can tell about $4 million in funding. They’re part of 500 startups batch 20, and now they serve trucking fleets as well as OEM. I’m not sure how they serve both of those groups. But I’m excited to find out likewise, I think that we have some really cool themes that are going to thread through upside and specifically through this interview, one we already touched on which is the industry the other one which you touched on as well is this idea of transitioning from technical co founder to CEO which we’ve now had a couple of conversations about.

Jay Clouse 5:08
If you guys have any thoughts on this interview as we go through, please tweet at us at upside FM or email us hello@upside.fm

Eric Hornung 5:15
And now for some validation

Jay Clouse 5:20
you know, Eric, when I was growing up I did not have cable

Eric Hornung 5:23
You didn’t have cable? why not?

Jay Clouse 5:25
literally had only your basic channels, the fox and CBS, the NBC. My parents didn’t believe in giving that to me. And so one of the biggest changes in my life was getting cable and then Netflix. Netflix will change the game.

Eric Hornung 5:40
Are we supposed to have a pity party for younger Jay here or is this where we goin? \

Jay Clouse 5:45
Most of this podcast is built to give Jay a pity party but not right now what I’m talking about is the idea of being able to binge TV and the fact that Netflix brought that to the surface and now podcasts are doing the same.

And that’s what Nick Oberhouse said on his iTunes review, “binge worthy.” I listened to a handful of entrepreneurial focus podcast. And up until this point, I’ve been consistently jumping around from podcast to podcasts in search of good content. In comparison, I just realized that I’ve listened to 14 consecutive upside podcast since finding the show.

Eric Hornung 6:19
Man. That’s a heck of a review. That sounds like something that I’d want to read live on our applied podcast.

Jay Clouse 6:25
That sounds like something that I like to read every morning when I wake up.

Eric Hornung 6:28
Wow, look at us waking up reading great podcast reviews. If you’re a listener, and you want to leave a great podcast review, maybe five stars throw review up on Apple, and we might read your podcast review live here

Jay Clouse 6:47
Ken welcome to the show.

Ken Sills 6:49
Well, thanks for having me. Very excited to chat.

Eric Hornung 6:52
We’re excited to have you here. We’d like to start upside with a question on the background of the founders. So can you tell us about the history of Ken?

Ken Sills 7:01
Ah, the history of Ken. Ken grew up in a place called Windsor, Ontario in Canada which is effectively a southern if you can believe it. Southern suburb of Detroit, Michigan. So I grew up in basically effectively the Motor City My father was a fleet maintenance manager so he worked in garages with big trucks all his life I put myself through university by working in those places and early jobs from I think the age of 14 on working in garages so I have a lot of experience with the vehicles of the type that we actually work on. My father was an award winning fleet maintenance manager is very good at his job and approach to effectively like a scientist but he wanted me out of that industry in the worst way. So at one point I You caught me reading his diesel mechanic books and literally took them in government, like you will not be a decent mechanic. So I encouraged me to do and my brother to get out and do something very different that why Oh, get educated.

Jay Clouse 7:58
Why was that?

Ken Sills 7:58
it’s a hard job. A very hard job so you know as a diesel mechanic you’re working in the shop strange hours. Very hard, dangerous job. Sometimes he would do things that were that he regrets now I mean, especially back in the 70s, when he was working on the shop floor, some of the things you were doing would be dangerous for your health dangerous for your hearing hurt your back lifting engines though it’s a hard job. And he graduated over years, started off as a mechanic and moved up to fleet maintenance manager which it gets you out the shop floor and like that is very good at it. But simultaneously, it’s also a very hard job in that vehicle breakdown. They break down and in a very frequently and the problem with that is there’s one guy who had when it hits the fan, that’s who gets the call. I remember being a kid and we’d be out in the Detroit River on our little boat of water skiing or something like that. And his people would go off and he’s like packet and boys and that happens a lot.

It made him happy, right? So he’s just like, go get an education, get out of this, you know, do something that that does a real best job, right. And we like that he encouraged with my brother and I to do that. And my brother turned into a severe weather scientist for Environment Canada. He’s the guy who basically chases tornadoes, and tells everybody would damage they did just fun. And I got out and did computational astrophysics. So I have graduate degrees in in computational astrophysics, which means that I did computer modeling simulations of what it’s like in the insides of stars and made predictions about things like cosmology, like how the universe formed using those types of models. So I was really early Big Data computer geek before the word big data were formed. I was using Linux operating systems to to tie together distributed systems to solve extremely difficult problems on large data sets and then comparing it to observable on

The stars that are available also worked fairly extensively in astronomical instrumentation, which is effectively bridging engineering with this physics side of things and building the, the camera, the devices that go on the back end of telescopes. So I have a lot of experience in developing hardware or research labs and the light and I’ve cameras, a mite that I decided to build on research telescopes, like large telescopes all around the world

Eric Hornung 10:28
\Before we head too much down the astronomy path, because I know I’m going to end up there. You mentioned that in the 70s there was it was a very dangerous job, this diesel mechanic job. What’s that job like today? Is it still has it evolved? Has it changed? Has anything changed about it?

Ken Sills 10:44
Boy has it ever changed. I mean, it’s very interesting question and that it has changed so much. The job of diesel mechanic right now is extremely difficult in a very different way and it’s about who is suited to do the job at this point, it’s effectively become so difficult because of the computer subsystems that are being put on these vehicles are getting very, very complex. And in fact, what we need an engineer, you know, a computer engineer and electrical engineer to be working on them. So diagnostics have become extremely difficult. The action of doing an oil change that hasn’t changed, right? A lot of the very basic stuff if you need someone to turn a wrench, there’s a lot of people out there who are good at turning wrenches. Diagnostic mechanics right now, there are 40,000 unfilled positions in the US currently and it’s expected to be 400,000 in the next 10 years and it’s just you cannot pump out the specialist fast enough. So that’s actually a problem that we address head on is how to make better and more efficient use of a really really tight resource which is these diagnostic mechanics so yeah, it’s changed a lot It is a little safer there’s no doubt back in the day my dad breaks that they would be using had these massive springs and I’m a massive I mean like incredibly high amount of force packed into a small thing called the magic break. And his job at one point used to be literally cutting those open basically they’d be in a candidate got a big cage and you break them open so that they would explode and like blow a lot of danger, basically. And that’s also he used to have to do repairs on the inside of trailers. And you’d be riveting with no your Pretecktion riveting on the inside of a trailer you can guess how loud that would be when you’re in a closed environment all made of aluminum and your riveting so yeah, it’s it was not a great thing. And I’m sure that that I hope that that’s imperfect people take safety a lot more serious, you know. But that’s, that is still a really hard back raping, breaking job. But on top of that, it’s something that requires extreme expertise with the tools that are out there today, which look a heck of a lot like a laptop.

Jay Clouse 12:47
So as a kid, what drew you to start down that path to the point that your dad had to say, No, get off that path. What was attracted to you about the industry?

Ken Sills 12:56
the thing that attracted about the industry is more the turns out now can look back on it now realize it’s science, like what my dad was doing with science, they might have called it like being a mechanic or, or even with him fleet maintenance management. But everything that he approached, he approached in a scientific manner, we break things up into pieces and really solve them in an intelligent way. So, you know, I was always interested in science and engineering. And effectively what he was doing, though it was called something else with science and engineering. It’s all about how you approach the problem. So I saw the way he approached problems. That is really cool. I want to do that do and the obvious thing when you’re when you’re growing up, right across from us, right? Is the obvious things, you gotta do cars, you can do drugs, it’s what they do.

Eric Hornung 13:34
So you’re going to do cars, you’re going to do trucks, and then all of a sudden you’re doing astrophysics. How did that switch happen? That seems like a little different.

Ken Sills 13:43
Yeah, it’s actually it is is not the is not a path that I was expecting, either. be quite honest. So high school up my eyes. I was honestly, I had no idea that I was how to tell this story.

I always thought I was dumb. My brother’s smart. He’s two years older than me, but just two years old. Me from a very young age, I grew to be the same size as him. So everybody thought we were twins, except that was two years younger. And when you’re or five or six years old, someone’s two years older than you, you’re pretty dumb.

I grew up thinking I was really stupid. And I just never could understand things that my brother could just easily pick up. And it took until I was probably I don’t know, like late grade school. So, you know, grade sixth grade seven. Before I started to recognize that I was actually able to do stuff that other people weren’t. It’s just helped me out. My brother Dave could do it. But, you know, other people weren’t able to do it. So I only started to really even understand that I could do interesting stuff. A little later, when I went to high school, I started to get exposed to a much deeper math and sciences and really was interested in it. It was really good at it. I remember in dollar americans i should translate. Great. 910, 1112. What is great 12 here, Senior?

Jay Clouse 15:00
senior. Yep.

Ken Sills 15:01
there you go. I remember my senior. I’m a Canadian, just the other day in my we just count.

But in my senior year of high school, I remember in my physics classes, the mark on the test would be whatever I got. And strangely enough, people didn’t want to kill me. And I don’t know how old that one off. But the rest of class knew this and would still encourage me to do well. And I looking back, I’m like, I did not get enough.

Jay Clouse 15:27
You were the curve breaker.

Ken Sills 15:28
I yeah. So they just said, you know, he’s not allowed to get above 100%. So whatever you got on the test will make it up. And then everyone else will be out of that mark. So I knew I was good at that. But the strange thing is, in high school, I was not destined to be a scientist, or an engineer, or any of that I was gonna be a musician. All my friends were musicians. I did music and I was going to do it professionally. I was a trumpet player, and I was very good at it. And that was my thing. And I was heading to university for that. But in my senior year of school, I had wisdom teeth removed. And for a few years, I lost all feeling and I lower lip because they, they hit nerves. Wow. And so that was the end of that you can’t play jump, but you can’t feel your lip. You can’t play trumpet that simple. So I, I then was like, Okay, what am I doing with my life. And it was as simple as opening, you know, the little books that they send you or I guess you’d maybe adult back in the day, my day at books, and you read the book and the bookstore, do what programs were available on like, at a university and astrophysics is one of the first things that you open up, you see, so that’s how I ended up doing that.

Jay Clouse 16:39
Tell me more about when that happened. I’m sure it wasn’t just okay. Well, now, what am I gonna do with my life? I’m sure that had to have been tough. What was what was that? Like?

Ken Sills 16:46
Not really.

Jay Clouse 16:47

Ken Sills 16:47
I wish I could say yes, it’s true. It sounds like it should be. But I don’t know why I just, I pivoted off of that really gracefully, and that it was just like, no, that’s not going to happen. A lot of my friends actually are professional musicians. There are people that I can grew up with. And I’ve gone on to do that as a life. I enjoy sharing that wisdom. And I don’t feel like I totally missed the boat on managing the different like that like that I’m doing I’m quite happy with as well. I actually have the, like, semi professional afterward and bands and stuff and enjoy that while I was a professor.

Jay Clouse 17:18
Yeah. So you skip over things like anthropology, you still didn’t you know, you didn’t pick the first thing in the book. So what was it about astrophysics that you said, you have? This is where I’m stopping in the alphabet.

Ken Sills 17:28
As I said, it’s, I’ve always been good at science. And I always think like a scientist. It’s like, it’s an obvious path for me, I looked out. And I’m like, Yeah, I mean, we’re talking about something that sounds really cool. It’s big questions, you’re asking, you know, asking and answering really huge questions. And I like that I really liked the idea of thinking big thinking out of the box, and trying to tackle him really hard problems. And I had done enough schooling at that point, to recognize that astrophysics, something that would allow me to exercise that person in my brain. So it was seemed obvious to me at that point, that it was, okay, let’s go for this, it was a real challenge really go down that path, I had a very, very strong High School. And that sounds like a great thing. But it turns out, it actually is a little dangerous. So I know I entered into university and was very, very well prepared in the first year to the point where I really didn’t need to develop any study habits whatsoever. And when you’re a big fish in a little pond, in other words, you’re you’re very much more knowledgeable about what you’re taking, then all the general population, you can get away with that in second year university, when all of a sudden you have a group of 10 people who are majors in astrophysics. And they’re all pretty smart. And you’ve been this actor who did not develop at all any study habits whatsoever, that’s a second year it’s painful.

Eric Hornung 18:46
You mentioned that you left the world of astrophysics, can you tell us about that decision, because kind of got into it, you said that you made a lot of stuff, you worked on some telescopes, you did a lot, and then now you’re out of it. So kind of walk me through the process of leaving.

Ken Sills 19:04
So it was just like, in and then out. So I went off and I did undergraduate degree and a graduate degree in astrophysics. So master’s degree, and I was I was doing research then on what are called non radio stations and stars. So you examine how vibrations in the outer atmosphere on the conductive layer will cause effectively noise, I mean, it’s very noisy process, you can make a jet engine held out that is on the outside of a lot of stars, that’s what you’re getting the music of the cosmos. Right, exactly. Except you can’t hear it. Because there’s a lot of nothing between us, there’s nothing to transfer the sound energy to us. But that sets up resonant waves inside a three dimensional wet resonant waves, much like if you and I were holding a skipping rope were shaking it up and down, there are certain patterns that form that are stable, you get those three dimensional patterns around the surface star, and you can use that to infer what the density profile is in places that you is literally impossible to see inside of a start. So that’s the kind of work I was doing research on in my master’s. After that, I decided I want to get into industry more recognize it was brought up in Detroit. So it’s, it was it’s always been one of these things where it’s like, I love being an academic. And then after a few years of it, it’s like, I can’t take this anymore, Can I do something real. So I bugged out and went off and work in new him. And I listened to him Can I work in East Hartford, Connecticut, at a company called Advanced fuel research, which did was a small business that did innovative research. And primarily, I was working on two things, which is continuous emission tomography, which is effectively mapping of flames. So you’d be you’d be looking at things like the effluent from coal burning plants, or gas burning plants that making energy or optionally burning waste product, and turning it into energy to optimize based on what the chemical signature is of the absolute exhaust. So I did that. And I also did a type of X ray spectroscopy that was used for identifying how heavy metal contaminants in do Egan God Superfund sites, these are places that the government is polluted beyond direct mission, so many billion dollar problem. And they enlisted us to create techniques to identify how those chemicals are bonded to the soil, so they can design more efficient cleaning solutions for getting that or making it not get into the ecosystem effectively. So I bugged out and I worked for that company did some excellent research there, that was a heck of a lot of fun. And we were building product, right, it was a real neat thing, because it’s like, in the end, you’re building a spectrometer, you’re you know, you’re, you’re delivering a solution. And I did that for several years. And then I was moved back into doing a PhD of Astronautical instrumentation at Ohio State University, where I worked on the some of the stuff for the large binary telescope, but also another program where we were designing a set of instruments that we continuously monitor certain types of stars, so that you’d be seeing them in multi channel so infrared simultaneously with optical around the world at all times. So you had to make sure you had multiple telescopes so that as the earth spine, you’d always be able to be tracking it and seeing it in real time, I worked on a lot of the back end systems that were used for that deployed that

after I finished doing so that was a really interesting thing, because they that that tied in, so adventure research, I did a lot of engineering work, where I was really building hardware and digging into the software that was running the hardware. So I was kind of an embedded software engineer at that point, plus data science, right? So everything that I do always kind of pivots on data science, but it’s like, what tool do you use to get data

Jay Clouse 22:27
And it’s the late 90s, I’m sure software development looked a lot different than it does today.

Ken Sills 22:33
Well, I guess it looked a little different. I’ll be honest, he, I don’t think anything’s changed ever. Software, software. And programming is programming. And if you if you asked me what language I use today, the answer is, it doesn’t really matter or what I shouldn’t say what developing environment, I use a particular developing environment called them, which is kind of esoteric, but that’s our whole shot actually uses been, we’re a little odd and that way, because that’s not it’s not pretty, it looks like it was great in that sense, nice as it was, but it’s still a very, very powerful tool that a lot of people who are who do take it seriously enjoy working anyway. So but like, like, which language it is that you’re right, it does change how you tie into things does change, but effectively programming is programming and hardware is hardware. And you’re you’re pushing electron from here to there to figure out which flip so so at some level, nothing’s really changed.

Jay Clouse 23:23
So you’re at Ohio State at the time, and then you actually so you’re at Advanced fuel research, and then you decided to go and back into research in academia. And then it seems after that you went back into industry, with Ericsson. So you kind of have this, yeah, this ping pong ping of, you know, like you said, academia to industry, academia to industry, at some point, you’ve kind of, you know, today, you’ve returned to this trucking industry with Bri tact, help us close the gap there and tell us how that came about.

Ken Sills 23:54
Yeah, so I worked as a professor at a research university in Canada for was nine years all told, and the last few of those years, every year, at the end of it, I would go into the department chair and say, Okay, I’m resigning this year, I’m going to go off and do something else because, but my intent was no more than three to five years. Being a prophet is just, I like going back to my my roots and doing industry stuff. At the end of five years, there was this constant back and forth. Okay, I’m moving on now. And they’re like, Well, how about if we do this for you, but we can make this easier, we can make this map into your life. I had young kids at the time. So like a flexibility of being able to schedule things around, you know, when I had to pick up my daughter sounds like that was very it was coming. It was good job. They treated me like gold. I gave me carte blanche to design courses from scratch, and gave me really interesting courses to work with. Historically, I’m a bit of a showman I literally have gone on stage with a rock band. And they are 30,000 people. I know how to have fun with an audience. And they wanted to bring that into the classroom. Mars, they gave me literally carte blanche to take one course that they were teaching. That was for first year physicists, if you will, that are not as as, as they hate physics. They never wanted to take this, but they’re being forced to. And courses get evaluated by at the end every year. And students tell you how much it sucks. And this one was a failing grade, every year everyone hated it, it would literally get a failing grade from students. Unfortunately, it was the largest course. So every year this is a training people to hate this x. And those people go out and they make decisions in the world, like, Hey, we fund physics. So not really wise to let that continue. So they said, You know what, go out there have a good teachers, I can actually make people understand things, but find a way to connect with these students. And within one year, we had raised that from our our failing grade, worst course we ever taught to our top scoring course know, changing outcomes. People still understood exactly what they understood before. But it was being driven to them in a way that really resonated with them. It was I really enjoyed that. But at the at the end of it, I started getting the itch again and saying I needed to do something I honestly didn’t start, like, the way I was expecting it to this was pre tech started as a small weekend type project that I was just going to have some fun, I have graduate degrees and electrical Computer Engineering. So I know how to make hardware. I understand data science, and someone came to me with an idea for for something that would allow you to hook into the port on your car that would take the data from the car when it has a check engine, like one would take that data and then send information to a dealership that would say, hey, you should order this part and schedule you know, three hours for this particular thing. And that was an interesting to me because that really was more app development, not data science. So I said no, but that really co founder came to me again and again. And again. He was very pushy which is good that’s what you want to really go founder anyways. Uh. That’s Peter bassa. And he kept on pushing me to think about that this. And eventually I started I thought about it enough that I realized that there was a data science problem that was unsolved. And it was really interesting because I started acquiring data off that just to get a sense of what was going on, on that computer plus the data bus that’s, you know, a vehicle and I started realizing like you’d Google is I got it, well, someone else is taking this data and doing data science need Google, Google, you try to find as a whole is doing this, and no one was doing it. And it boggles my mind that there was a huge untapped data source, and it was just that data was evaporating. You can’t on evaporate data, you can go backwards in time and say, Oh, we should have been taking that. So let’s let’s just have taken that right, though, is I see, I finally said to that Peter bass, I was like, you know, what if, if you’re interested in pivoting this company into a data science company down for doing this. And so we started working on that early concept, the end of 2014, just taking some initial data, doing some initial prognostic on it on vehicles, my car, his cars, and the day families that thing and very quickly started to have success on the data that was coming off of that.

Jay Clouse 28:06
I just wanted to add a quick aside, I see you have a 4.5 on rate my professors. So it does send that you made that accessible for the students

Ken Sills 28:13
and some chili peppers or did they remove those now?

Eric Hornung 28:17
I think they’re gone now. Unfortunately, used to have them the chili pepper. Oh, yeah,

Ken Sills 28:21
that was that’s hurtful.

Eric Hornung 28:25
How do you define data science? I feel like there’s so many of these words that are getting thrown around right now. Like data science machine learning AI, I feel like everyone has them in some pitch book. So how do you define data science, and then we can kind of move on from there.

Ken Sills 28:38
Yeah, I defined data science, this almost the same way I would define science, it’s effectively like we’re taking data. And like, very specifically, what was the document talking about taking data off of car, right. And what the car has is typically a couple hundred sensors that are measuring temperatures and pressures, torques vibrations, at times, all around the vehicle. And it’s a bunch of disparate data points. But you’re trying to make a picture, right, you’re trying to make a model and the second year trying to turn the sensors into a sense of what’s going on a context that’s doing science. And if you’re using data, that’s data science. So like, I have a very loose definition of data science, because to me, I’ve been doing data science for very long time, I, all of the positions that I’ve ever worked on, are using data to make inferences about something in the real world. And that’s data science. And, you know, initially, we would call that producing a physical model these days, you use physical models as well, statistical models, but also this cool new thing called machine learning models, which are just statistics, let’s not get confused, that everyone’s using these these terms, like their magic they’re not, it’s just a really, really interesting way of building a statistical model that can take in that ton of data and make sense of it in a way that a human would have a real hard time visualizing. And that’s what we do. So I never saw is an AI company. Because artificial intelligence to me is there’s a very specific definition of what artificial intelligence is. And that’s about creating an autonomous intelligence that can think like a human and interact with you in such a way that you believe you’re addressing, right? Human, we use machine learning, which is a type of statistical technique that will allow us to take a lot of data and from that data and inputs, but labels as to what the outputs of those data’s when it all happens together. What does that mean? Do you label it in that way, you can learn from those patterns.

Jay Clouse 30:34
So how would you describe what Preteckt does today

Ken Sills 30:38
Preteckt delivers vehicle diagnostics as a service to only EMS and fleets. So what that basically means is we predict when check engine lights are going to go on, sometimes months in advance, we do that using raw sensor data, and it’s not threshold. So recognize when a vehicle built these days, whether it’s a small vehicle of it makes no difference, the OEM the manufacturer, the vehicle specs out all the sub components separately, like they’re talking to one company to make a starter, and another company to build a power steering pump, another one to do the AC system, right. And then all those things come together and get squished into a box and sent on the road, those interactions are where we come in, we understand holistically how they come together, and how they interact with each other. So that when you have a subsystem manufacturer whose job it is to understand a power steering them deeply understand that power steering pump, but all that they can say, is OK, if it goes at a speck of it’s too hot, or too cold. Or if it is, you know, the pressure is here, over here or over there. So, here’s my box for my power steering pump. But unless you know what the rest of the vehicles doing, and unless, you know, okay, but how about when it’s doing this, right? If it’s that idol, that box changes, right at idle, it’s tiny little box over here. Whereas if it’s a dump truck, and it has it’s, you know, it’s lifting as data, it’s up here. So they have to make a big box reach subsystem that are all the possible ways that this thing can behave before they throw a diagnostic trouble code, check engine light, right? But what we are able to do is define these little boxes based on all of the other boxes simultaneously, all the different parameters come together and form a picture of the context of the vehicle. We provide that context we make sensitive, how is that integrated into the vehicle to start to they have to put that into the vehicle when they manufacture it? Or is that something that you can plug in later? Can you get it over the air? The answer is yes to all those things. So it depends on how you’re who you’re working with. And we’re working with a manufacturer like a General Motors, for instance, they have their own ability to acquire data so that their own path and Justin in jams case that’s on Star, that data would come into their cloud. And they could use our back end that we’ve developed our architecture that supports the machine learning, they could use that to provide better vehicle health management to their customers, so they don’t the input and the output side with our fleet customers, some of our customers, they will use our hardware, we actually have hardware that you can install directly onto the vehicle in under a few minutes, very easy. So they install the hardware, and then it pumps that data into our backend via the cloud, and then go straight up to a web dashboard that goes there feeding this manager. So it depends on the customer, some customers, we have end to end the customers is just the core of our technology. And then somehow one end and not the other.

Eric Hornung 33:22
You mentioned that you guys can predict in some cases, months in advance a check engine light coming on. What else? And this is a lot of data. So what what else can you predict? And what else? What are the other outcomes? I guess,

Ken Sills 33:36
I guess what else would you want me to predict we are focused in life is making sure that there is no downtime, I guess there’s Chico’s it. So it’s interesting, we entered into it, our vision going into this was that we were going to eliminate unplanned downtime for vehicles. And we can’t, for the most part, keep a vehicle for breaking, they’re going to get old, they’re going to break. But what we can do is tell you when it’s going to happen. So you can be prepared that and get ahead of it, and not have it break down on the side of the road, right, it’s like this will break down three weeks, as is driving by the next time we’re going to fix that part, make sure it doesn’t break down. That was our original thesis. And and we do that. And we do that across many different subsystems on a vehicle. So whether it’s being the charging system, you know, your cylinders, the after treatment system, the emissions control system, all these different things are something that we’re simultaneously covering. But it turns out that is not just the fact that we’re preventing downtime, that we’re catching this beforehand, and therefore, you know, saving them both the road call, but also the downstream damage. When one part breaks, oftentimes, there’s a chain reactions of things that get out of whack, cause damage downstream. So we save a lot of money on that. But it turns out that one of the one of the other benefits that we have is decreasing the amount of diagnostic time that it takes a fleet to evaluate what the problem is, and act on it in in the garage. And that’s actually a really interesting part of our machine learning strategy. And it wasn’t where we started, when I started the company, my mentally I was like, machine learning that’s going to hit the home run, right? It’s a black box, we can just use these things, and it’s going to magically solve all of our problems wonderfully. And a lot of people are still believing that’s the case. And we learned it’s not the first year, two years of our rd, if you will, where we’re taking data and really designing that back end to support prognostic, we had the idea that if you had enough data, and you had the right models, you could just predict everything with 100% certainty. And it it doesn’t work with the current state of the art. And it’s not just us we know it doesn’t work oftentimes I’ll talk to there was one truck manufacturer who got on the phone with that top of them on and off for years now. He’s the head of the innovation department there, he you got on the phone with me is I’m going to tell you this is going to be a very aggressive call. I’ve had so many people blowing smoke up my ass over the last us about this ml stuff, solving it. And none of it works like yeah, you’re right, it doesn’t. So you know, here’s here’s where we’re different. And you know, number one, we take boatloads of data no one else has taken that kind of data and honestly, garbage and equals garbage out. That’s a fundamental tenet of science, right? So you have to take great data. And so we entered this as a data science program, right? Like, to me, I’m a data scientist, first thing was how do I get great data. And since we couldn’t find anybody who had a piece of hardware that would give us great data, like we’re making that piece of hardware. Now, we do not want to make that hardware that is not our dream leaves on the back end, right. But you’ve got to have great data. So that’s one way that we differentiate ourselves. But the other way is, is really about how we how we learned machine learning. So I guess that when we started out, we had this linear path that generate data, do machine learning output to the customer. And the problem is, you can either have too much to the cotton to the customer, because what you’re catching is multi variant anomalies, which would, you’re saying, you know, here’s 1000 sensors and some weird going on with one of them doesn’t fit the rest of your out of pocket, right. And if you send that to your customer, I’ll often going to be wrong with a fleet customer, if you sent off to a commercial fleet. And they’re like, Okay, well, there’s probably pulling it that just cost them money. And if you pull it in there, like they’re not wrong with this vehicle, you just burn that bridge and last customer, they’ll maybe let you get away with one or two of those. But it’s a big deal. So then what people are doing is optimizing to reduce the number of false positives, right? Unfortunately, that means you’re missing all kinds of stuff. So there’s no real winning right now with that strategy. So what we’re doing is we have a two stage process for the first stage is we use the multivariate anomaly detection using physical models, statistical models, and machine learning to put those anomalies in front of a human. So we have our own expert who reviews using our data tools, all the historical data for that vehicle, and all vehicles like it and has access to the real time data on the vehicle right now, the performance tests that they need to do just like it’s shot. So we have all those potential false positives, going through a human having that human inject their intelligence into it, and make the like, provide a label, right, so they’re going to label it as cautionary tale is it useless is this is how to ignore this is it cautionary as it’s gonna be a problem in a few months, you know, hi, severe critical or Yeah, I mean, this is about a breakdown, right, so they label it for severity, and also do a root cause analysis. So if it’s a problem, like a severe critical problem, and like, and this is the part and this is why it’s that part right, with all the data to back it up a two stage process. First one is multi nominally texting, the second is human in the loop, and we have a whole separate second section of machine learning that’s dedicated to watching how that human interacts with the data to improve the efficiency of that humans interaction so that we get our KPI which is number of vehicles per technician way up into the stratosphere,

Jay Clouse 38:43
is that human that is checking that is that a Preteckt human? Or is that someone at the shop checking the truck and you have tools for them to put that tag on and send it to you and you trust them to do that

Ken Sills 38:54
Great question and the answer is both so we have our own human that checks all this stuff and and jack says we’re really codifying his brain jammin and machine learning which is terrific because he’s really good but we also work with partners to have their experience come in. So when we’re working with OEM for instance or or tier one suppliers they understand let’s say their power steering pump better than anybody we have a great relationship with Bridgestone for instance, we’ve been working with them for a while and we have a multi year contract with them. Now they understand tires deeply better than we do. There’s no doubt they have expertise in tires, we have expertise in contextualize in the tires, and really understanding, you know, what makes a good predictive model for where and failure so we work with them, they inject their expertise into our back end. And we learn from that and train get the models better. So depends on which customer

Eric Hornung 39:42
do the models change based on the type of vehicle?

Ken Sills 39:45
the models do change based on the type of vehicle, you have to take that into account as one of the variables that said, you’d be surprised how similar they are we working with the Memphis Area Transit Authority, which is the first bus that we installed on, we only work with long haul vehicles up to that point. So we’re working on a very different use case vehicles utility class, and we warned the Memphis Area Transit Authority matter that chances are, it’s going to take us a few months to train models, because we’ve never we’ve never done it, but no idea what buses are going to look like. But they’re like, we need you to solve this problem. So let’s let’s go within the first week, we had predicted something that was going to fail, they brought it in, and they’re like, yep, you got it. So we want them it very quickly. Because we started we were using models for vehicles are completely different. But there’s enough similarities in the context of how a vehicle is being used. And certain subsystems very quickly, you’re able to make value for customer out of what you’ve already learned

Eric Hornung 40:43
the hardware that you mentioned earlier, that hardware specialized to clean the debt, I didn’t really understand the link between the hardware and getting great data.

Ken Sills 40:53
So on a vehicle, there’s a network, the a local internet, if you will, that connects together a whole bunch of engine computer units, each engine computing and it controls a subsystem. So for instance, the emissions control system might have several easy to use, and you probably units that each one of those has several sensors hanging off that and is using that information and transmitting and back and forth between other components on that network to make sure that the whole vehicle is operating together simultaneously. So what we do is we tap into that that local area network and both sniff and pull, so but for the most part sniff, we take the data that’s already been chatted around there, and send that to the cloud. So quality data Don’t even think about as pre processing. It’s how many of those messages are you actually grabbing, it’s a lot of messages. So we’re taking between one and five gigabytes of data impressed per vehicle per month, when you’re looking at a standard telematics solution. It’s about 1000 times less than that, because they’re only there looking at, okay, did a check engine light just pop on, if so, let’s take 30 seconds. And you know, before and after that, and kind of take that little chunk of data and send it off. But to us, if you want to predict something more than 30 seconds in advance, need the data that happened more than 30 seconds in advance. So we’re taking all that data constantly, permanently and storing it, it’s been a very successful technique for us to be able to get precursor data to failures.

Eric Hornung 42:11
How many vehicles is the Preteckt system on now?

Ken Sills 42:15
the hardware is installed on hundreds of vehicles, but our backend has been used on 10s of thousands of vehicles.

Eric Hornung 42:21
And when your back end is used, are you still storing all of that data that one to five gigabytes per vehicle per month,

Ken Sills 42:28
it depends, when the back end is used, we’re using typically data that has been ingested from another source, like a telematics service provider. So some of them have very poor quality data. And that makes it harder to use. And some of them have good quality data, which makes it easier to use. The nice thing is, we’ve seen we’ve seen things with laser vision, we’ve seen extremely good picture of what is actually happening. So after you’ve seen that picture, it makes it so that you can now look at someone else’s poor picture and go oh, yeah, but I know what that means, right? But you have to have the clear picture first, before you can build a model that can work for data.

Jay Clouse 43:02
So I get why OEM is probably didn’t have this to start. They’re probably excited about making the data and maybe didn’t have the experience of the prioritization to start predicting it in house. Is there a threat to pre tax of OEM saying, we’re going to do this analytics within our own systems in the coming future? Or is that actually kind of an exit strategy for you guys?

Ken Sills 43:23
It’s both, you know, there’s always a threat, you have to recognize that it’s not just us that feels the threat, though it’s every OEM, everyone right now is fearful for their life of being disrupted of existence tomorrow, you see massive companies right now, pivoting, making huge pivots, look at General Motors, right, really deciding we’re not doing that anymore, we’re doing this, that’s a big company to be making those kind of decisions. So disruptions in the air, everyone’s scared, and that actually works for and against us. I mean, it does mean that someone some OEM is undoubtedly working on this. But what we’ve found is they’re looking at us and saying, Yeah, but we need to win this and we need to it yesterday and your four years ahead. So how about how about we work together on this. So we’ve had a lot of interest for Halloween, more actually more interest that we can manage right now. So we’ve actually gotten to the point now where we have to hire dramatically in order to support what’s in our pipeline of customers coming in, such as OEMs.

Jay Clouse 44:17
yeah, it seems like you’re kind of in a picks and shovels situation with autonomous vehicles, where you’re, I’m guessing your hardware and software does not become obsolete, when vehicles become autonomous.

Ken Sills 44:28
Actually, we love autonomy, because one of the primary modes of feedback or maintenance right now on a vehicle is the driver, the driver is like, something’s not right here. doesn’t feel right. I don’t have the power I used to, there’s a vibration, there’s a noise smells, ROM. Hey, it’s on fire. As soon as you go to autonomous, you can literally like imagine for you for a moment, if you don’t have a driver, you don’t have to stop the vehicle, right? So you could load something up at a port on the east coast in Maine, and drive it continuously all the way to LA, right? Do you don’t have to, because you don’t have a driver restriction on number of hours. And it’s very easy to put your fuel tanks on it. So it could be leaving main have a problem and be on fire by the time it’s in Vermont. And you don’t know, right? So the neat thing about autonomy is that as we get closer to solving that problem, what we’ve seen is as we are like, Oh, hey, we’ve almost saw this, this, this is a real thing this is going to happen. They’re like, we haven’t even thought about how to do predictive diagnostics on it. And we’re removing the driver. So all of a sudden, there’s a bunch of red flags going up going, Oh, hey, does this so in the last in the last year, we’ve had a lot of inbound interest from OEM who are the ones that are close to solving this problem because they recognize that they have to have solved predictive diagnostics before they can launch autonomous vehicles

Eric Hornung 45:47
when you say oh I just like to clarify who that is. Is that you mentioned GM a couple times it seems like we’re talking about like the 18 wheelers that you see on the highway Are you talking about more than that are talking about like Peterbilt, Navstar, who is this?

Ken Sills 46:04
so yeah OEMs. Cummins would be a type of OEM and that they make a an engine they don’t have a an actual vehicle itself but so Cummins would be an OEM in my mind but obviously Navistar another one Freightliner would be another one that’s on by dialer we work with nine lived through Thomas book classes which is also owned by dialer General Motors is no end of General Motors does not make heavy duty vehicles. So we work with them on their passenger vehicles that so you can you can map out the Nissan super Toyota Ford those are all we EMS and then you know if you’re getting on the other side of the space when you’re getting into heavier duty vehicles were talking about Freightliner and Volvo and can work and pack higher so those are all OEM we will call Bridgestone, even though there are a manufacturer, the manufacturer a subsystem effectively, and so in that way, we will call them a tier one supplier.

Eric Hornung 46:56
On the other side of the customers. You mentioned fleets, how would you define and that that’s a little more of a fragments and space, right.

Ken Sills 47:03
So fleet space is really interesting. And that so our definition of a fleet is, hey, you own more than one vehicle as a fleet. And interestingly enough, it’s a long tail problem on the short side, and that most fleets are small fleets. So the majority of fleets are 20 vehicles, and under and you know you you get a lot of vehicles in Super fleets, but most vehicles that are in fleets are not in Super fleets. We as a company are targeting the large fleets like fleets that are having a 200 and up vehicles. So that that’s the area that we’re targeting. Because the cost of sales makes sense for us, we can also target 200 and lower through going through providing our back end service to telematics service providers who already have figured out the the, the path to market, they have a channel for selling to smaller leads. So all they have to do so almost every telematics service provider and in North America, all of those of you all commercial vehicles are connected this point. And legally, they have to before the ELD the electronic logging devices. So there’s now a pathway to data that almost always has the ability to do an over the air update to change what data is being sent to the cloud. So we can enter into a partnership with any ELD manufacturer provider, as well as any telematics provider to say, okay, you want to deliver this type of prognostic to your customer send us you know, these 15 variables sample that these different frequencies and, you know, ingest that into our back end, and we can send you a flag that says, hey, this is going to be a problem in two weeks, and you can send that to your customer. So when we’re targeting smaller things, talking it through channels, such as that.

Eric Hornung 48:37
So we’ve talked a lot about the different types of customers and the different types of vehicles they have, you mentioned earlier, that the buses was something that you had to kind of on ramp to, are there other areas that you would need to on ramp into right now, like heavy duty caterpillars type stuff, or smaller, like sedans, like are those different enough that you would need that on ramp time as well are those even target at all.

Ken Sills 49:00
They’re currently not targets, but they are targets for the future. So that we’ve had a lot of interest from passenger vehicle manufacturers as well. And passenger vehicle fleets, rental car companies, for instance, we’ve had a lot of interest from heavy duty manufacturers, everything from the cranes, there’s an awful lot of interest in in that space, anything that has a lot of downtime. So if you look at mining equipment downtime is exceedingly expensive. And it’s effectively the same kind of thing. It was a different use case.

Jay Clouse 49:26
Ken you mentioned a couple of customers one being super fleets that sounds like you’re saying your hardware to the second being telematics company that sounds like you’re selling the software to are those are two main customer segments at this point.

Ken Sills 49:39
Add into that the manufacturers So actually, most of our revenue right now is driven by manufacturer. So large contracts with with we said to my suppliers,

Jay Clouse 49:49
Can you describe that model, then with those three different types of customers

Ken Sills 49:54
I have to be careful about describing models with those types of customers, because they’re much more sensitive to us, describing them all with those types of customers. But recognize that when you’re working with a manufacturer, they’re trying to improve the way that they’re, they’re delivering predictions to their end customers, when it when you’re talking about someone who would make a subsystem like a battery, for instance, they would be perhaps thinking about how to move their service from selling a battery to someone and sell it set of that selling the service of a battery to someone so that they know I will always have a battery that’s going to work in my vehicle. And it just cost me $5 a month, I no longer think about the fact that I have to go get one every three months, or Hey, I’m driving in these conditions. So we work with manufacturers to two and two and suppliers to really help them understand when is something going to fail? How do you model your business around selling that as a service instead, and then give them the tools that they would need to be able to make sure that things aren’t failing for those customers who they’re providing service to.

Jay Clouse 50:55
So how do you think about the market ahead of Preteckt them? Because it sounds like any one of those three customer segments are pretty large markets in and of themselves. So help me quantify what this could look like if pretext just goes gangbusters.

Ken Sills 51:08
Yeah, so the path in the near term. So 2019, we’re really focused on getting there’s a particular KPI that will set us on fire, if you will, which is a number of vehicles per technician that are being used in our back end. So this human in the loop sand industry standard is that you’ll have about 25 vehicles per technician in your fleet. We’re already up above 250 vehicles per technician, which means that we’re by that’s great. We want to be up above 1000, before we start pushing into the commercial vehicles said like massive fleets and skis and writers. You have great unit economics at that point. So it makes a lot of sense for us to really launch big time into those fleets. The way to get there is to do great data science. And strangely enough, the best data science is not done with the commercial fleets. Because a commercial fleet, you tell them, there’s a problem, they will go out and fix it that day,

which makes it hard to data science because you don’t get to see what actually happened. We love working with municipal fleets. And for the same reason that municipal fleets have a really hard time which is there a resource constrained and never have the resources to act on everything we send them so they can only act on the highest priority and then we send them so we send them a cautionary we know it’s going to get ignored. And now we get to track it and say that cautionary got works, you know, it is now high and that cautionary, it got worse. Again, it’s not sit here, we really need to be talking about fixing this. So that as a data science accelerator is terrific. So in 2019, we are really pushing hard on the municipal side going into municipal fleets, primarily public transit. We also work on school lessons, for instance, you have to recognize that the vehicles themselves even though there they are buses, the engines of power units for those are designed not to work in those type of use utility cases, right, the design for long haul transport is shoehorned inside of a bus. So they also fail more frequently. But in the same ways like that. So they expand contextual space, if you will, and we really understand deeply, a lot more about the prognostic and then we can use that to improve the number of vehicles that we’re handling per technician. So our path and 2019 is to work with a few top notch commercial fleets while simultaneously growing on the municipal side to really ramp up that KPI 2020 will be launching into like scaling into the commercial fleets. And the whole while we’re pumping, pumping the gas here and, and really making sure that we’re doing well, by maintaining our and growing our relationships with a bunch of Oh, he ends in Tier One suppliers, because everyone wants to know that you’re working with everyone, like, what would we provide context like? So the OEM want to know that you have fleet data, and you have great relationships in the fleet side so that they can understand what how the vehicles are being used. And the fleets are like, oh, oh, you work with the experts. You’re on this subsystem. So yeah, I’m having a problem with that subsystem. What, let’s talk about that. So you really have to kind of solve the problem simultaneously.

Eric Hornung 53:59
What is the pricing model for Preteckt? Is it per vehicle per month, SaaS type fee? Is it How does it How does it work? And does it change on a customer by customer basis?

Ken Sills 54:10
it’s a per vehicle per month. And it does change based on whether or not you’re using our hardware, or you’re providing your own data stream, or even if you like, if you own the customer, right. So if it’s telematics service provider, and they’re only using our back end, and they’re doing API, injection of data and API, output of the product, Gnostics, that really reduces the cost of us providing that service. So for for that type of, you know, where they’re just using our backend engine, that’s a licensing fee,

Eric Hornung 54:35
help me understand the costs a little bit, it sounds like data storage is going to be a huge part of it, you obviously have some really smart people on your team. So I’m guessing those salaries are big. But what else? What else is driving cost?

Ken Sills 54:47
data storage actually drives almost zero.

Eric Hornung 54:49

Ken Sills 54:49
yes, we really think of that as a zero, we store everything, and it’s effectively free. If you start start sensibly, it’s effectively free. So we really kind of turned the whole thing on it head. And instead of looking at what the costs are, we primarily focused on the value that we’re delivering, and trying to improve that dramatically. So we just completed a multi month Those are my former PLC pilot project with a major commercial transport company and showed on their own internal analysis, which is important on those just kind of spit balling and saying, Hey, we did this, and we worked with them, we sent them all the data, all the prognostic data, their own deep dive analysis showed that we were saving them on average almost $3,000 per vehicle per year. So to us the fact that it costs us around $15 per month for the data to sell you a connection so this is all real times were transporting an hour or so that’s a lot of money and that that is so much money that a typical telematics service provider that would break their model for delivery right but for us being the ones who are really finding that back end and understand the product Gnostics were concentrating on the big side so that we can go off to other partners such as service providers are always and say, Look, this is how much money is sitting out there untapped. Give us these chunks. And we can deliver this fraction at a team. So for data costs on our side, if they’re using our hardware hardware is expensive hardware. But yeah, I mean, you have to amortize that over the contract. And that’s, that’s an expensive piece of that data costs or expenses, because we do send a lot of data are there we do expect to be optimizing that substantially using both losses and lossy techniques over the course of the next year, to drive the margins a little bit better. But we’re a small team still can people we have to be very, very careful with where we’re assigning those people, right? Because you’re right, they’re expensive, smart people. And I can either say, hey, develop a better model that’ll take that $3,000 $5,000,

or I can say, hey, the same smart person take my cost down from $15, $12.

And I know which one of those two things I’d rather have them focus on. So we’ve always we’ve always ignored margins. And look at that big picture sideways, like, we can create boatloads of value

Jay Clouse 56:54
at $3,000 per vehicle per year saved what is like, how big is the pie, how big is what you can say, for these companies?

Ken Sills 57:03
Absolutely. So the for the municipal fleets in North America. And actually, I should say, I should qualify that as Canada and the US the total addressable market does not mean as simple as 500 million. The total addressable market in the USA and Canada for commercial fleets, is over 10 times that. So $5 billion. As soon as you go into the global market, obviously, you’re getting multiples on that rather dramatically. And we are starting to go into the market, we are working with companies that are in Japan and China are ready to bring this to market there. So there’s a lot of value that’s sitting out there waiting to be tapped.

Jay Clouse 57:36
So can I saw on your LinkedIn that, you know, you built the prototype, from what I can tell, and you were the CTO at the start, and now you’re the CEO. What was that transition

Ken Sills 57:45
they transitioned occurred in the end of 2016, beginning of 2017, what we found was that our early stage CEO, who is really, really good at getting everyone excited and onboard very good at establishing relationships with my fleets, we got to the point where we’re starting to speak to a lot of large fleets that we were driving jet like a big interest in so large fleets, and all yams. So when you’re talking to them, they’re expecting to talk to the person who really deeply understands the product, you have to be able to communicate to their very excellent engineers, why you’re solving get them, they weren’t a so what we found was, it was it was required for, for me to be bridging from CTO into SEO, because I was the person who really could communicate the vision for what the company can do, and how it can do it. Fundamentally, I came into this having never done a business before at this my first startup, when we first moved down to Memphis, Tennessee, our companies headquartered and started our first business accelerator, which is start go, I was a bit of a snowflake kind of difficult to deal with, in that I said, very specifically, the rules or engagement are very simple. I will work on this, so long as I never once have to speak to a customer or an investor, and if you make me do that, I quit.

So that I would fire me right now. But it is what it is. So it took a long time for me to get comfortable to the point where as an academic, I recognize what sales actually was when you’re doing it, right. And sales right now is me connecting with people who are really excited about the thing that we’re making. And I can do that all day and all night because I love making and I know what it can do. And so I’m talking to people who get it and are like, I want that thing. How do we do it? I love that process. So when I realized that there are guys that sales is not a selling someone something that they don’t want, when it’s Hey, I made this for you to solve your problem. And you have a big problem, buddy, I bridge that gap. And around that same time that we were moving into larger customers where I was talking to people who understood that they needed this product was around the same time that I was like, Okay, yeah, I can

Eric Hornung 59:48
Did that cause any internal conflict.

Ken Sills 59:50
I mean, we did change CEOs and our initial co founder is with us anymore. So the answer to that is a cause them in our internal reorganization. So we’re still working with that co founder, he is still you as equity. And he’s very much cheerleader for us. But yeah, I mean, it causes significant reorganization at that point. But we all understood that for the company to grow to that next phase it we had to reorganize, we had to get on the right path.

Eric Hornung 1:00:18
Talk to me about the decision to headquarter in Memphis, you’re Canadian, you were up there as a professor right before this. And now you’re down in Memphis, Tennessee, it just I don’t have the I don’t have the Natural Bridge.

Ken Sills 1:00:32
So the Natural Bridge is start start. CO is a business accelerator. That’s, that’s based in Downtown Memphis, Tennessee. And they have there’s two reasons for Memphis number one, it’s a logistics hub. It’s the very center of the states. So you have all the East Coast stuff coming into the middle and all the West Coast stuff coming to the middle of huge intermodal for shipping. So huge, huge center for international shipping. So their access to customers, if you want to talk about the number of fleets that we can touch within 20 minutes, it’s insane. So we were we were sold on this, but we were up in the Toronto Canada area developing this, all three founders are from that area and start co found us and said, Hey, we’re interested in this product. And we have a market for you. We were looking at passenger vehicles, they’re like, ignore that you don’t want to be doing that we have a much deeper pain for you to solve. And it’s in the commercial trucking space. And they have ties into they were able to introduce us to that their network to to give us access to those early customers who had to take a big bet on us. Because when we came down here, we really had very little right, you need data to do data science. And we have never been installed on a truck. Now you have to get someone who’s buying on spec potential for this system. And circle really helped us with that

Jay Clouse 1:01:46
we had Katie on the show recently. I think that’s, you know, that’s who introduced us. That’s, that’s wild to me that they found you in the early stages of what you’re doing. Do you have any insight into how they were able to find you at the early stage and match make you to what they had access to in Memphis?

Ken Sills 1:02:02
Yeah, so we were part of a Google Next entrepreneurship program in the Toronto area, and the managing director of star co at the time, Mara Lewis knew about that program, and was always out scoping things. So they’re very good here. They know they’re in the middle of the country, and that they know, they know, they have to go places and really find talent. So they’re always shopping at anything. That’s a real interesting event where, you know, extra, you know, interesting startups are coming to so they’re out there with the ear to the ground, they found us there and they immediately engaged us and said, Hey, you know, this is we can do this, we can make this work for you. We haven’t regretted it, it’s been very successful for us

Eric Hornung 1:02:36
is there any piece of data that is not currently collected on vehicles that you think would be incredibly insightful to have like, Is there a sensor that is not currently best practice staff

Ken Sills 1:02:51
Yes, but the thing is, it’s already coming It’s so close. So we’re there’s one company I’m not going to name names here there are companies right now that are just releasing products that are going to solve that and what it is is vibration analysis on the wheel hub. So if you look at where you connect your wheel to not the tire tire goes on we’ll we’ll do is we’ll have putting a vibration sensor there that’s connected to a connected device so it doesn’t have to directly to connect itself but something that would like let’s say Bluetooth or Wi Fi into a hub like ours let’s say and then take that data off to the cloud that is the holy grail you can do so much with that vibration data because and that’s stuff that is not currently on the canvas this this local area network that connects everything all the running gear is separate and it breaks down constantly so this is very exciting for us in that we’re working with Bridgestone on tire we’re models and we know full well that that vibration data has value from from rubber all the way up through you know the brakes wheel hubs themselves because when we will have sale that is a 500 kilograms or thousand pound pay Hello, that’s being curled in the space at very high speeds. very deadly. And this happens so you know, you can predict when that will help skills all the way up the actual you bolts transmissions, you get a whole new window on that. So yeah, that’s coming down the pipeline. And it’s big news. It’s really going to make a difference to what can be predictively modeled. This has been awesome. can last question for me. Your dad works so hard to get you out of the trucking space. And here you are back in it. How do you feel about that? Now it’s funny because I’m in it. And I’m not in it. I don’t really know. It was a really hilarious thing that I moved down to Memphis Tennessee and I think it was in June of 2015 I had Preteckt unit. I want our first one that we’re installing it a truck on product united hand standing in a garage and Memphis Tennessee truck garage. And I took a picture a little selfie and send it to my dad say, How did I get here because it looked just like his garage. standing behind me was a truck a look just like one of his trucks. And it’s like, I’m like back in. So yeah, it was a it’s a big, big twist simultaneously. It’s, it’s an interesting twist on it. And that, you know, and I love it. It’s actually it’s, it’s, as I said, My dad’s good. She’s really, really good at what he does. He is very, very intelligent scientist when it comes to what’s going on in an engine. And there are still times and this is the last time I really called on and was within the last year where we’ll get a really strange set of data, just like what is going on here. And when when, when it really comes down to it. And then I need help. And like, there’s just no one else to call it. He’s retired and honestly, I don’t when his mind gets worrying. He just dives in and so I but every once a while, I’m like, Okay, let me share my screen with you. This is what we’re seeing as hills spinning his wheels for about 15 minutes. And then they’ll come back to me and like I’ve got it is what it is. And but then he then he’s so excited that for the next week. I’m like, I also thought about this. And this is like your return calm down. Go record. Some music is also a musician.

Jay Clouse 1:06:03
That’s awesome. Well, thank you so much for being on the show. Can if people want to learn more about you or about pretext after the show? Where should they go?

Ken Sills 1:06:09
www.preteckt.com that’s PR e te C k t.com?

Jay Clouse 1:06:18
All right, Eric, we just spoke with Ken sills of free text, or do you want to start? Or do you want this deal memo you want to talk about the opportunity you want to talk about can as a founder,

Eric Hornung 1:06:26
I think we should take this back to a more classic upside deal memo. And let’s start with the founder. Let’s start with Ken. I was impressed. That’s my first take. I’m I’m going to ask you for yours. I’m going to take the floor here. I was very impressed with Ken. From a intellectual standpoint, you could tell that he had been converted almost to this idea as well, which I really enjoyed. He didn’t come on board thinking that this was going to be something that was going to be the next 10 years of his life. And once he got in to the actual problem. He was like, all right now I’ve been converted, which I think is like very disproving a hypothesis, scientific method kind of stuff, which I really like. And just overall I was I was very impressed by Ken, what do you think,

Jay Clouse 1:07:12
totally agree in that I love when people understand something so well, that they can take a very complex idea and break it into very simple terms, so that anybody can understand. And throughout the interview, Ken was super, super good at that he was able to read our reactions and, you know, sort of physical responses on the video and tailor the way he was explaining pretty complex things to us in such a simple way that you have to really, really understand something to be able to do that he’s obviously incredibly smart. His background in astrophysics, mostly just out of interest in science and curiosity, you know, that bleeds through into what he’s doing here with Preteckt, as you said, didn’t seem like he was necessarily looking to join a company and really do this thing, but was convinced otherwise had the stipulation of okay, as long as they don’t need to talk to a single customer or investor, but in talking with can, he’s got to be the person doing that. I mean, he’s, he really understands the product, he understands the problem, he has a background in the industry, and he’s able to get that across to I think just about anybody. And the way that he explains it.

Eric Hornung 1:08:22
what do you think about this idea that they don’t use the standard software’s the standard languages, he said that they used a product called them?

Jay Clouse 1:08:31
Yeah, I forget exactly what vim is, or how it works. It’s not really the language. It’s a way of doing coding. When I’ve worked with developers in the past, and my product management background, we were not using them. But we were using I think you mentioned Python. And we use that quite a bit because that is much easier to program from a statistical and mathematical standpoint. So I love when a founder comes on, it says, I built this prototype. And you know, it sounds like Ken has been teaching himself how to code for long time, the way that he approached machine learning and data science. He says, you know, data science is just assistant statistics. This is actually kind of a pattern we heard at CES this year, too, is people kind of the pendulum swing the other way from AI and machine learning. And big data is just like cool, big buzzwords, but moving towards the practical application of, Okay, if we have these things, what does it actually mean for the end user and how we make that useful? And it’s clear that he’s on that side of the the argument as well.

Eric Hornung 1:09:29
Yeah, the reason I bring up then is, it sounds like it’s something that isn’t very widely known, which brings up a bit of a shadow and a bit of a opportunity in my head. If you bring on some engineering talent, and they’re used to Python or rails or whatever else, I don’t really even know the competitors of them. But how easy is it to learn it, he mentioned it was a bit archaic, or it was a bit I forget what the word he used, does that make it harder to recruit people in place that it’s already kind of hard to recruit people. The thing that’s beneficial about using a non mainstream technology is that you have a competitive advantage over your competitors. Amazon famously did all of their internal communications on their own, they built out their own infrastructure, and it sucked for a while, and people hated it. But because of that, they were able to figure out things that sucked about hosting websites, and messaging and all the other problems that led to them creating AWS. So I think that there are some shadows about building on a niche system and some potential opportunities.

Jay Clouse 1:10:43
I believe VIM is just where people are doing the actual writing of code. So it’s, you could do this in a text document. A lot of times, a lot of people use sublime because it will color code certain actions that you’re doing, it makes it easier to look at, had a page of code and see like which operations are working, which are not, I think them is just an old old school version of that. So I wouldn’t be worried that people don’t know how to use them. I think most people who have moved off of them, just like the alternatives and the way they display information, and it’s not really changing the core of what’s being done in that environment.

Eric Hornung 1:11:20
Okay, so then all of my takes are cold, we’re throwing them out the window, but that’s okay. Because I still got to get a point across. It doesn’t apply here.

Jay Clouse 1:11:31
It’s all good. I mean, I just when he said that, I knew I’ve seen them. And I’ve seen people talk about writing in them, but I knew it wasn’t like a language. So assuming we’re on board here with Ken as a founder. Let’s talk about the opportunity ahead of pre tax. It seems like we’ve we’ve spent a little bit of time here in our episode with super dispatch. This is a little bit different. What did you think about the opportunity here?

Eric Hornung 1:11:56
So I worked in business development, which is essentially in internal mergers and acquisitions. I know that sometimes people call sales, business development, sometimes they call market research, business development. But in the job I was doing it was essentially internal mergers and acquisitions, business development for a company in Cleveland called Bendix commercial vehicle systems. They make the brakes that stop these big trucks. So Freightliner Volvo all of the ones that he mentioned. That’s what they do. They make the air brakes. In my research, I did a market analysis of all the different companies that are out there that make any part that goes into a big truck, there are hundreds, Jay and a lot of those make over a billion dollars in revenue a year, like this market is massive. And people don’t really know about it. Like, these are companies that are in Sheffield, Ohio, and Aurora, Ohio, and some random town in Michigan and some random town in Indiana, and some random town in Pennsylvania. But they hire me couple hundred engineers. And they are just churning out this one specific product that goes into every single commercial vehicle system in the world. And it’s crazy because there’s so much money in it. I did a quick look, just to see if I can figure out okay, how big is this diagnostics market. And the first number I found for the Automotive Diagnostics scan tools market for both passenger and commercial is 52.73 billion.

Jay Clouse 1:13:32
Whoa, that’s crazy. He didn’t give us any numbers that were that big in the interview. I mean, he talked about fleet sizes, I wonder where that 52 plus billion dollar expenses come from, in terms of the actual customers, because the customers you mentioned were super fleets. telematics companies and manufacturers. So I guess he gave us the numbers for fleets, we didn’t get numbers from the telematics companies, which sounds like a lot about what you’re what you’re saying there with the 52 billion or manufacturers, which I would expect to be a lot of money from, from either of those. So yeah, I mean, I love this interview. And you’ve broken down our market size buckets before per se.

Eric Hornung 1:14:12
yeah, so there’s four market buckets, there’s less than a billion dollars of market size, that’s what we would consider a pretty small market that you have to dominate, there’s one to 5 billion, which is a decently sized market. But you still have to kind of dominate, there’s five to call it 10 to 15 billion. And that’s a good sized market on a yearly basis, that’s a healthy market, you don’t have to dominate. But you do have to, you can, you can ride the wave of that market. And then there’s everything that’s bigger than that. And once you get bigger than that, it’s just it’s very big. And I don’t think there’s a point for an early stage investor to fight over a number, that’s 25 billion, or 26 billion, the number doesn’t really matter at that point, it’s the fact that it’s a large enough market think that this market is a large enough market. And the extra data provided by Preteckt the extra data it’s pulling is only going to increase that.

Jay Clouse 1:15:12
This starts to get to where I get a little bit fuzzy in this interview, because we start talking about Ken and his background, I’m like, this is awesome, this is awesome, starts talking about the technology and how they implement it, and this talent on their team to figure out how to use it in such a practical useful way for customers. And I’m saying this is awesome that he starts laying out these customers. And then I start questioning things like how many people are on the team, how are you fulfilling all these different things that are going on at once we have a partnership with Bridgestone we’re talking about we’re working with the Memphis Area Transit Authority, we’re working with super fleets, and also telematics companies, and also manufacturers, we’re doing software, and we’re doing hardware, we have three different business models, you know, with the SAS and the hardware. So I just didn’t have enough time to, like, pull everything together and really understand little more tightly, how these things are all working, because I got really excited about astrophysics, and some of the earlier parts of the interview. So you know, my, my gut tells me, this is huge and Preteckt is doing something that no one else that I could find is really doing. And that’s going to be super valuable. And I think it’s extremely valuable, even in an autonomous age. Because as Ken mentioned, I don’t think those companies are necessarily thinking about the maintenance that goes on with these these trucks or even cars that are autonomous driving across the country. I think he’s in a really good defensible position here, I just ultimately couldn’t come to a place where I’m like, Okay, this, this opportunity is this big, it just kind of got to a place in my head where I thought huge, and almost maybe too good to be true. But I think that just comes from me not having enough time to ask more questions.

Eric Hornung 1:16:49
I feel like there’s one core product and they expanded kind of ancillary services based on because they have such huge customers. So when a customer says, Hey, I need hardware, and you’re like, well, we didn’t really want to do hardware. But that’s a huge revenue number, I think that you expand the hardware. One thing that I really liked about the business model you’re talking about their team is this idea that they are trying to eliminate both type one and type two error. So they have the machine learning algorithm, which is doing its thing. And then they have a person who’s kind of reviewing the machine learning algorithm. So it’s not just one layer, okay, we’re going to send this truck to the shop, and they gets the shop, and there’s nothing wrong with it. And it’s not the other way where, okay, there’s an alert, but it doesn’t seem that bad, let’s not worry about it. And then there’s downtime. So I like that there’s kind of this two layer approach to making sure that when they send something to the shop, it should be being sent to the shop.

Jay Clouse 1:17:50
Yeah, everything you said just spoke to, we have systems in place to make sure that we are continuously improving on what we do best, which is the data science there’s a lot of Pretecktions around it even seem a little as you’re saying, a little human intensive to make sure that there’s not only a technician on behalf of the company or somebody who’s fixing the vehicle there but that Preteckt is aware of and analyzing what’s happening so they can improve the system’s themselves. And when you mentioned that the number of vehicles per technician being used right now, industry standard is 25 vehicles per technician Preteckt has gotten that up to 250 vehicles per technician and wants to get to around 1000, before going into commercial vehicles. That speaks to a huge cost savings in my mind that pretext should be able to carve quite a bit of revenue out of

Eric Hornung 1:18:42
did I understand this, right, that they only have one technician right now?

Jay Clouse 1:18:45
I did not have that note. So I’m not sure.

Eric Hornung 1:18:47
Okay, let’s assume that I am right. I think I remember him talking about how incredible this technician was. And maybe it’s a master technician, and there’s a couple under him. But to the extent that there’s a concentration of employment around this one person, that’s a shadow for me, because this person has learned your system better than anyone else. He’s the one who can transmit and if he leaves or something happens, then that number, that 200 number that they want to get to 1000 might get reset back with someone new. So there’s definitely an employee consolidation risk there.

Jay Clouse 1:19:25
Yeah, I hear you. And I think you could be right, you mentioned that his entire team is 10 people. So there’s probably one individual as overseeing all the technicians that are working on the vehicles, he’s not the one that’s actively working on it, he just kind of monitoring to my understanding what is going on when the vehicle does get a signal to go in for repair and making sure that it’s consistent with what they’re reporting. But I hear what you’re saying. And I think you’re right.

Eric Hornung 1:19:48
But I guess that’s not different than any startup where there’s employee consolidation. I just think that generally in a startup, the on ramp time for a new CFO or a new cmo you take the general skills you knew and you kind of bring them over this is a proprietary system that is continuously being developed so not only are the computers learning but this technician is learning as well over time so I just think it takes a little bit longer to get back up to speed should anything go wrong

Jay Clouse 1:20:17
this is our second company out of Memphis following on from some of AK He also spoke highly of Memphis Memphis being a logistics hub with access to customers within are so many customers within 20 minutes he said was huge. All of their founders were in Toronto and they started and start co found them and brought them into the Memphis market one I’m impressed with how Stark I was able to kind of find a company that really fits the niche of Memphis convince them to move down and have such an impact to this continues to fall on our theme of Memphis being a pretty collaborative place granted cannon Preteckt were also introduced to us by Katie at start Co. But it’s it’s fun to get more and more data points in one ecosystem here to point to what’s going on.

Eric Hornung 1:21:08
Yeah, just broadening that out of that I love and we get three to four data points in a city. And we can kind of start to understand that ecosystem from a couple different perspectives. And the theme of collaboration is definitely one that has popped up and stuck in Memphis Speaking of which, I’m going to be in Memphis in November. So I will will have to swing by and report back

Jay Clouse 1:21:29
Do you have family there?

Eric Hornung 1:21:31
I do have family there.

I do. I’m going to a wedding in Brownsville, Tennessee. So I’m going to spend Thursday night and Memphis get me some barbecue.

Jay Clouse 1:21:43
All right, Eric. So six to 18 months from now what are you looking for, from Preteckt, to be even more excited about this opportunity,

Eric Hornung 1:21:52
I want to see that they’ve expanded a few more vehicle types they said that there is a bit of an on ramp process when they went from commercial two buses I see this obviously getting into the passenger market eventually but that’s not what I’m talking about. I’m talking about maybe mining equipment or which is a massive market or construction or some other type of heavy duty medium duty type expansion because I think that to me, this is going to be just a long sales process so I don’t know that 1618 months is going to be that indicative for their current customer base but if they can get some new trials and new spaces that’s going to lead to more and more sales over the next 18 to 36 months once they get some data on new potential verticals so I want to see some trials in new spaces maybe it’s more bus maybe it’s more busing like city trials like they are doing in Memphis maybe something like that but out of their kind of core niche to refine those algorithms so that they can attack different verticals that might have shorter sales funnels.

Jay Clouse 1:23:03
I’m looking to see the traction that they have with the the fleets and the types of vehicles that they’re in now how that is growing. But also Ken mentioned disruption is in the air and everybody is scared and thats related to autonomous so I’d love to hear where conversations are with autonomous customers whether it’s the OEM or telematics companies also working with autonomous customers and see what the future looks like for Preteckt in an autonomous world to see if they’re a market leader

Eric Hornung 1:23:34
Do you think that there’s going to be big moves in autonomous and the next 6 to 18 months?

Jay Clouse 1:23:39
I think so I mean every projection you here as people’s saying, you know, this is going to be on our streets and and ruling the streets and the next year, I think who talks like that a lot of you guys

Eric Hornung 1:23:49
There’s gonna be a rule in the streets!

Jay Clouse 1:23:53
But I hear some pretty aggressive projections for one autonomous vehicles are out on the roads in an hire mass. But in any case, I think behind the scenes for them to get ready to go into an environment that will probably be more regulated. These companies are going to be sharing up everything on their their back end and behind the scenes. And so I think that Preteckt is in that conversation. Awesome. Well, Jay, if people want to interact with this episode, where should they go? If you guys have thoughts on this episode, please tweet at us at upside. FM, or send us an email hello@upside.fm. And if you liked this episode, send it to a friend send it to a friend who’s in vehicles or repair or diesel mechanics. We’d love to hear their data science and machine learning,

Eric Hornung 1:24:38
man, we have a lot of stuff going on in this episode.

Jay Clouse 1:24:40
Yeah, send it to your mom. We’d love to hear her opinion, astrophysics. Yeah, especially if your mom is an astrophysicist. Let us know what you think. And we’ll talk to you next week.

That’s all for this week. Thanks for listening. We’d love to hear your thoughts on today’s guest. So shoot us an email at hello@upside.fm or find us on Twitter at upside at and we’ll be back here next week. At the same time talking to another founder and our quest to find upside outside of Silicon Valley. If you or someone you know would make a good guest for our show, please email us or find us on Twitter and let us know. And if you love our show, please leave us a review on iTunes. That goes a long way in helping us spread the word and continue to help bring high quality guests to the show Eric and I decided there are a couple things we wanted to share with you at the end of the podcast. And so here we go, Eric Hornung and Jay clouds are the founding partners of the upside podcast. At the time of this recording. We do not own equity or other financial interest in the companies which appear on this show. All opinions expressed by podcast participants are solely their own opinion and do not reflect the opinions of death. And Phelps LLC and its affiliates on your collective LLC and its affiliates or any entity which employs This podcast is for informational purposes only and should not be relied upon as a basis for investment decisions. We have not considered your specific financial situation nor provided any investment advice on the show. Thanks for listening and we’ll talk to you next week. Some way to relax.

Interview begins: 06:44
Debrief begins: 1:06:15

Ken Sills is the co-founder and CEO of Preteckt.

Since developing the initial prototype of this hardware/software solution in 2015, Ken has been focused on building Preteckt’s team of engineers, scientists, business developers, and sales professionals so that the automotive industry can benefit from Vehicle Prognostics as a Service. Ken has over two decades of R&D experience creating solutions that couple hardware, software, and data science.

Preteckt uses machine learning to predict vehicle service needs before they get serious. Trucks on their network learn from each other so that our customers can predict breakdowns weeks or months before they cause downtime. Their connected hardware gives real-time, unparalleled insight into vehicle maintenance needs. It’s like driving with a master mechanic in your engine compartment.

Preteckt was founded in 2015 and based in Memphis, Tennessee.

Learn more about Preteckt: https://preteckt.com/
Follow upside on Twitter: https://twitter.com/upsidefm