Ep. #860 - Protecting Your Time, Effort & Data

00:00.00
Matt Watson: And we’re back for another episode of Startup Hustle. This is Matt Watson, your host for the day. Today, we are talking to Kirk Marple with Unstruk Data. As most of you know, I’ve been a software developer [in] the last twenty years; and my last company was all about data. So [I’m] excited to have an interesting conversation today talking about data, and all the things you can do with data. I’m sure Kirk is going to tell us how to get more value out of all of our data. But first, I want to let everybody know about our sponsor today. Are you thinking about starting a business or expanding a current one? If you are, then it’s important to get it set up and maintained properly. That’s exactly what the folks at Universal Registered Agents do. LLCs, S-corp, C-corp, nonprofits—no problem. Learn more by clicking the Universal Registered Agents link in the show notes. Kirk, welcome to the show man. But before we start here, you said your background, you’re a software developer yourself. So love to hear a little bit about your background and how you came to start Unstruk Data.
00:56.80
Kirk Marple: Exactly. Thank you so much for having me. Yeah, excited to be here.
01:14.64
Kirk Marple: Yeah, for sure. It started out . . . undergrad and grad for computer science back in the day. It was really kind of learning about software across all kinds of different applications that I was working on. That happened to be in the media space.
01:34.43
Kirk Marple: Worked on image processing applications, early video streaming—really, it wasn’t intentional. But I kind of always centered around those kinds of applications. And then ended up going to Microsoft after my masters. That was a great experience and I started some software companies after that. But really started out purely as a dev, and kind of grew into management and took off from there.
02:00.30
Matt Watson: I saw you were from Seattle and I just assumed [you] had to work at Microsoft. Because, I mean, especially now that you’ve been there for a while. And in the industry, over the last few years, anybody who works and lives in Seattle has either worked at Amazon or Microsoft.
02:09.16
Kirk Marple: Yeah.
02:18.53
Matt Watson: Because they hire almost everybody there. Crazy, isn’t it? Like 20% of everybody who lives in Seattle works at Amazon or it’s like some stupid crazy statistic.
02:21.79
Kirk Marple: And they bounce back and forth too. That’s a funny thing. So, yeah.
02:30.91
Kirk Marple: It has to be a crazy number. It’s just insane. It’s grown. I thought Microsoft used to be the beast here. But yeah, now it’s Amazon.
02:36.87
Matt Watson: Yeah, so how did . . . who gave the idea to start Unstruk Data? Where [or] what was the genesis of that?
02:45.69
Kirk Marple: Let me take you back a few years. As I said, I have always kind of been in this media space. And then after Microsoft, I started a video transcoding company, so it was [in the] early days of web video. It kind of evolved into more broadcast focus video, so like every PBS station in the country had software that I wrote. It’s kind of interesting because it wasn’t even intentional. One of my first jobs out of college was working on image file formats. We were writing parsers for TIFF files and fax files and stuff like that.
03:19.11
Matt Watson: That’s super nerdy.
03:25.40
Kirk Marple: But it’s a weird thread that I’ve always had, through my career, [to deal] with file formats and parsing and metadata. I’ve never gotten away with it. But it’s been a bit of a sweet spot. So that idea is not too different from the video transcoded company. It’s really broadening it out. Now, we deal with video-audio documents, 3D files, geometry—all that kind of stuff. So it’s a data platform but for unstructured data—any file-based data.
03:56.93
Matt Watson: Whoa.
04:00.70
Kirk Marple: It’s kind of a parallel universe to the SQL kind of data platform.
04:04.24
Matt Watson: For our listeners, let’s talk for a minute about what unstructured data is. Because I don’t think everybody realizes that I’m a software developer, so I get it. But can you explain a little more about what unstructured data means.
04:17.17
Kirk Marple: Structured data is typically rows and columns, like a spreadsheet. It’s like your Google Sheet. There’s a structure to it, which is typically the rows and columns of unstructured data. Anything else that can live in a file, it could be log data, images; it could be documents. it’s almost this huge bucket of everything but SQL.
04:45.56
Matt Watson: And the simplest example of that is a bunch of PDF files, right? Think about [taking] all the invoices that come out of the electric company and we got all these PDF files. We need to do something with them or whatever, right? Like you have files that have text in them.
05:02.84
Kirk Marple: And it’s almost missing. It’s almost a misnomer too, because there is a structure in a lot of these files. I mean, if you write a parser, there has to be a schema or a file format. But the tricky part is [that] there’s a million of them.
05:05.56
Matt Watson: But they’re not databases with a pattern.
05:21.61
Kirk Marple: You have to know how to read a TIFF file versus an MP4 versus a PDF. And that’s where the trickiness comes in. That’s something I happened to randomly get a lot of experience with in my career. I keep seeing people, when they try to deal with this type of data, they always start soup to nuts. It’s like having to write a compiler just to build your program. And what we’re doing is just trying to take that heavy weight of dealing with this type of data and abstract away the parts that’s common. So I’ve always been a platform guy—always finding those patterns, that kind of 80-20 rule. What’s the 80% we can do that you can do once? And then refactoring [it]. So what we’re trying to do for the data is let people plug in the unique parts, and let us take the messy 80% out of it.
06:12.78
Matt Watson: So did you have a specific use case that you are focused on when you decide to start this or that? Like, “Oh, I have a buddy that is doing this specific thing. And I’m going to start a company and help them with this exact use case.”
06:18.77
Kirk Marple: It was a bit broader. I think it’s been four or five companies. We’re trying to build a platform like this to deal with [unstructured] data in the drone space, in the sports data space, like computer vision for sports. I was at General Motors working on data from the cruise vehicles. There was this format called ROS bag format, which is from the robotics format. Everybody’s DIY-ing this whole problem and this whole process. After I was at GM, I was like I could build this. I could take all my learnings from the past. Why couldn’t I do something like a media management system for any kind of media in the industry? It’s very different where the media is not for our eyeballs to get on to Netflix. It’s media that typically ends up in some sort of machine learning algorithm. But you also need to visualize it, like preview it—search on it and all that kind of stuff. So that was the first time I was thinking of starting this company. It was about five years ago when I left GM, but then I kind of bounced through a bunch of CTO jobs. But it really came back around where I tried to build it. We started building it at three or four different companies. I was like, “God, why does this have to be there? There has to be value here. Somebody needs to go do this.” So I raised a seed round about a year ago. Being a developer, I started writing the code for it probably about four years ago. I was thinking of building a podcast discovery platform where it would take podcasts and run audio transcription on them. Then sort of spider the entities, people, places, and things that are talked about in the podcast to create a kind of web that was searchable. So I had actually built that; spent a couple of years. Nights of weekends building that, and it became the IP that I put into this company. I had all this knowledge graph, data ingestion, and data enrichment stuff already built. We had enough to get going. I was able to grab a team of folks that I’d worked with before, mostly frontend developers since that’s a part I hate. That was kind of able to complement and start building this out for real.
08:39.98
Matt Watson: So how was it raising capital? Did you have any other founders/partners when you started this or just you?
08:47.34
Kirk Marple: There are two founding engineers and myself. They are really good frontend engineers. I’ve built frontend in the past, but I’m more of a backend guy. We grabbed 3 other people at the time, they were folks we had worked with before. I’m our head of QA and Ops and UX. We just hit the ground running and we had already been working at a previous company together. Few of us left that company and started putting the band back together. Basically.
09:20.54
Matt Watson: So what was the process like with raising capital? Was that a nightmare? And was that in Seattle?
09:23.69
Kirk Marple: I was in Los Angeles at the time, so we just moved back to Seattle. We’ve been away for a few years. It was during COVID, so it was all on Zoom. I think it democratized the whole process that it was probably easier.
09:42.36
Kirk Marple: I was working for a company in the Bay. Commuting from LA to the Bay before COVID; started raising money virtually. It was interesting. I bootstrapped my last company, so I hadn’t raised before, but I had a great group of people. From the CEO of my previous company and connections I’ve made over the years, we’re able to put the piece together. I don’t know if it was easier or harder than other people. It was definitely challenging. But it’s all kind of once you get one, it kind of triggers, and it’s this kind of building effect.
10:16.38
Matt Watson: Yeah.
10:21.12
Kirk Marple: We’re actually in the middle of a seed extension fundraising right now. It’s good to get feedback. But getting commitment and triggering and getting to that thing, we’re literally weeks away from all this closing right now.
10:38.53
Matt Watson: So for our listeners out there who are like you, entrepreneurs, thinking about raising capital. For those who are curious. What would you say? What percentage of your time do you end up, these days or the last few weeks, doing this? Is it 80% of your time dealing with this?
10:38.92
Kirk Marple: It’s always a struggle. It’s never easy.
10:51.20
Kirk Marple: But it’s interesting. It goes in waves and I’m classically more of I’ll-do-business-during-the-day and I’ll code at night. I’m more of a night owl, so I do all my meetings in the day. Have my team meetings [and] do fundraising stuff during the day. It goes in waves. I’ve met probably three dozen maybe or more investors in the last two months. They are usually half-hour calls, maybe four- to five-minute calls, and mean it adds another. Let’s just say another five to ten hours a week of calls for that. But I think it’s great. It’s amazing how much the story gets refined over a couple months. When we start thinking about it in December, you think you have your stuff together.
11:36.58
Matt Watson: Yeah, yeah.
11:44.43
Kirk Marple: You like, “Oh, I’ll get this down if the story plays out well.”Now, I look back and I’m like, “I don’t know how the hell we were even talking about it.” Because it is so much smoother after doing it 20 or 30 times. One big thing—we actually paid money to a firm to redo our investment. Which was some of the best money we ever spent. And even the one we kind of reuse, the one from last year. Now, if I look at the new one that we have, and we spent a lot of time on that, it actually was a big-time suck. But, man, the messaging then and driving it into marketing. You got to have that message really clean and tight. Our existing investors gave amazing feedback. I was like edit, edit, edit. And we have a tight clean story right now.
12:34.14
Matt Watson: For potential investors, you know, getting that pitch . . . it’s got to be super clean and clear and understandable. I think most people, they say way too much.
12:40.78
Kirk Marple: Yeah.
12:49.40
Matt Watson: As an investor, I need to know three things. Do I trust you? Do I understand this industry? Is this a big enough opportunity? Am I ever going to get my money back? They are pretty simple questions that everybody wants answered. And everything else almost doesn’t matter.
12:51.54
Kirk Marple: I mean . . .
13:00.69
Kirk Marple: Um, yeah.
13:04.50
Matt Watson: But people oversell so much complexity behind all of it.
13:08.95
Kirk Marple: It’s really true. I think I’m lucky that I’m a little older than the event. I’m not right out of Stanford and in my twenties anymore. At this point, they call it like a founder, market fit. If I have anything I have [it’s] because I’ve just happened to be doing this for the last twenty years. It’s always been on the spot. Yeah.
13:28.40
Matt Watson: Well, I don’t know how old you are. But from everything that I’ve ever heard, the average age of a founder for successful companies is about 45 years old. It’s not 20.
13:37.62
Kirk Marple: That’s what I see. I’m in my early fifties now, but it’s like I have already done a company for a dozen years. I have learnt from that . . . bootstrapping and then it was good. I got five years in the industry.
13:50.89
Matt Watson: Um, yeah.
13:57.30
Kirk Marple: Just being at other companies, but I was ready to do it again.
13:58.17
Matt Watson: The most successful founders are people like us. We’ve been around different industries for ten- or 20-plus years. We see the problems that need to be solved. And you’re like, “You know what? I can solve this problem and I’m just crazy enough to think that I can actually do it.”
14:09.60
Kirk Marple: You have mapped that risk tolerance. I think I weigh high on the risk tolerance skill. I think that’s the biggest thing . . . you’ve got to be a little crazy.
14:17.34
Matt Watson: I’m going to go start a company.
14:28.31
Kirk Marple: I mean, you can’t be too conservative. I did my first startup when my kids were little. Now, my youngest is 21, and he’s actually working for us as a QA engineer. I actually have more time now than I did my last.
14:37.99
Matt Watson: Nice.
14:43.51
Matt Watson: So, wait a second, his job is to find bugs in Dad’s code? Perfect. I love it.
14:46.17
Kirk Marple: And it is true, mostly. I don’t think he will be finding [bugs] in the frontend code. I don’t write my code. It’s perfect.
14:59.20
Matt Watson: I’m with you, man. I hate frontend code. This is a great story. I love it. So tell us about your customers today, and how they’re using your product. Like give us a case study.
15:08.88
Kirk Marple: One that’s really interesting is we talked to an aerial survey company. In December, we met them and they’re flying around . . . I mean, they’re capturing data for things, like the park service. [They are] doing things like wildlife analysis, and the problem they have is [that] they are technologically savvy in their own subject matter expertise, like they are researchers. But they are keeping their data on Sharepoint. They are keeping all this data they are capturing . . . treating it like office documents. But it’s like images . . . point clouds, all these kinds of other data; their concept of search. I’ve heard this even at Chevron . . . when you say search to them. They are like, “Oh, you mean, by file name?” That’s the only way that they are thinking. They are not thinking about semantics.
15:59.87
Matt Watson: Right.
16:03.39
Kirk Marple: What is this image? Where is it geospatially? Where is it in time? That’s what we really bring to the table. We’re creating an index that goes across all these different axes. It creates a tagging taxonomy and this knowledge graph. You can whittle down a big haystack into a smaller haystack. Do analytics on it. Get an alert and things like that. So [for] this customer, it’s a pure data management solution where they just need to become cloud native. Make their data more actionable, but also get “search” over a longer range of time. They typically think about things on a per-flight basis, not like, “Hey, show me all the dams I’ve flown over in the last three years.” To answer questions about your data, it’s manual. It’s the only thing they could really do and that’s really what we’re trying to step in. When we hear people that have [a] manual kind of clunky processes . . . [It’s] like a maintenance engineer looking for problems while they are walking around their plant.
16:50.98
Matt Watson: Right.
16:59.55
Kirk Marple: They are literally taking pictures with their mobile phone. They are printing them out and putting them on the wall to triage their data. And just by streamlining that, being like, “Hey, we can set up these triggers. We can set up computer vision algorithms to be, like “Oh, there’s water pooling.” There’s a crack, and then just alert you or your teams, just to compress that time.”
17:25.28
Matt Watson: With the . . .
17:32.72
Kirk Marple: One of the assumptions that I get, especially from a lot of investors, is that everybody is a snowflake. But it’s just a problem. Everybody’s problem is different.
17:37.46
Matt Watson: I was gonna say that there is a new wave of all this stuff. I think, partially, some of it is what you’re describing. Because of drones and stuff, right? At our company, Full Scale, we’ve done some work for some clients that do drones to take pictures of crops or construction sites and stuff.
17:48.92
Kirk Marple: Give exactly.
17:56.78
Kirk Marple: Yeah, that’s part of our . . . exactly.
17:57.14
Matt Watson: So, if you’re taking pictures of crops, you know what data it is. But if you can analyze the pictures, you know what percentage of it is green. What’s the health of the crop? And all these things that you can get. But then it’s like, “Okay, what do we do with that data later?”
18:10.60
Kirk Marple: Yeah. Right.
18:16.54
Matt Watson: Right? Then they could feed that data into your system [to] be able to chart out over time. It’s like, “Okay, on this date, it was this percentage green. Or this person [is] this percent healthy or whatever.” And chart all this stuff out over time; build some dashboards; build some intelligence on it; and then set up an alert, like you said.
18:31.13
Kirk Marple: Exactly. It’s the same problem with SQL data warehouses. It’s like we’re kind of warehousing this data. But then you [have] to make it actionable. You [have] to be able to do stuff with it, trigger it, and generate all the other sets of insights. That’s what we’re doing for the unstructured data side.
18:35.28
Matt Watson: There are a million different . . .
18:50.58
Kirk Marple: Dealing with documents is different than images. [Images are] different from point clouds. Having a canonicalized kind of system that you can bring anything into; and then get one kind of stream of data out that you can really act . . . that’s really their focus.
19:02.78
Matt Watson: So it sounds like the same complexity of setting up a new company. If you’re setting up a new business and maintaining compliance, it’s not easy. That is why it’s important to have experts help along the way. And that’s exactly what you’ll find when you visit http://universalregisteredagents.com for all your business set up and maintenance needs. They can help you set up an LLC corporation, nonprofit—wherever you are located. In addition to helping you create the right kind of entity, UniversalRegisteredAgents.com can also help you with Registered Agent Services. A wide variety of corporate services helping meet the needs of independent directors. I don’t like dealing with any of that kind of stuff, so having somebody to help me do that sounds awesome. So how often do you end up working with your clients? Kind of like on a consulting basis?
19:48.28
Kirk Marple: That is definitely true.
19:57.83
Matt Watson: Let me preface it by saying [that] data itself is awesome. But it is completely useless unless you can get some kind of insight out of it. It’s trying to get the signal from the noise. So, my last company, we did application performance monitoring for software.
20:07.56
Kirk Marple: Exactly.
20:15.68
Matt Watson: It was a data product. We took billions of data points a day, and it’s like, “Okay, so we got billions of data points today. What do we do with this?” As you mentioned earlier, actionable insights like useful information that I can actually go do something with.
20:32.50
Matt Watson: Do you end up spending a lot of time consulting with your potential customers to say, “Okay, you got some ideas, but we know a lot about this. And you can do 20 more things than you even thought about.” And you end up doing a lot of professional services and consulting too.
20:39.60
Kirk Marple: It’s an interesting point. We’re still early enough that we’re trying to make it more like a no-code kind of self-service environment. So we really don’t have to be involved at that level. But we’re actually looking to partner with people [with that] data.
20:52.56
Matt Watson: Sure.
21:02.99
Kirk Marple: Firms help build computer vision models, or connect up their alerts to their ecosystem. So there is a bit of that glue. One of the personas that we’re looking at right now is that service provider. A persona of how we can work well with the folks that are really touching the customer. So more of a B2B scenario. Our app is aimed more at a data analyst persona today, so it’s really no-code. No true ML knowledge or developer knowledge.
21:26.22
Matt Watson: Right.
21:38.92
Kirk Marple: Dump your data in and use it like Tableau, ThoughtSpot, or something like that. But we see this extensibility to the ML front. We can be a data prep platform to other ML tools. We can eventually just be like, “Hey, use us as a headless service through our APIs.” Somebody might be putting together a bigger solution set and they just want to drive us kind of headless through that. We’re trying to find people in that space who, rather than us, may be doing consulting. We would find a partner who would do that consulting for the customer, and then they could sort of build around us in that ecosystem.
22:17.59
Matt Watson: Yeah, I can see there being a lot of industry specific solutions. For example, it’s like, “Oh, we take photos of roofs. We have to analyze all these photos and figure out whether they have hail damage or not.”
22:25.35
Kirk Marple: Yeah.
22:31.44
Kirk Marple: Exactly. That was the company where I was the CTO previously. That was their forte—with the drones, it was like hail damage on roofs for insurance and things like that. But the thing is if you look at it, it’s just pure data management, data ingestion, [and] database storage search.
22:49.15
Kirk Marple: That has nothing to do [with you] if you’re looking at a roof or agriculture . . . At least some of them, their first initial reaction is that, “Isn’t this not generalizable enough? Isn’t everybody different?” And I’m like, “Look, it’s still a JPEG. It’s still an MP4 file; the metadata in the file is still standardized to some level. It’s the semantics of what you look at—the image—that’s different.” What it means is the ML models are really the unique part. So that’s what we’re trying to develop an ecosystem of. Look, we can get you really close and then plug [it] in, if you’re savvy enough. Here, go build one yourself and plug it back in. We have an auto-ML solution that we’re delivering that doesn’t really know anything other than just drawing bounding boxes. We can build it for you. We can basically hand off to other ML tools. Because that is what we see. That is where the secret sauce comes down [to] everything else. Once you do that and bring the data back to the system, it’s stored in this canonical fashion so you can search on it. No matter if you ran a computer model in Roboflow or Azure or AWS or Clarify, you could have all of these models running from different systems. And pull them back into a conical knowledge graph and that’s really the value we’re bringing.
24:11.51
Matt Watson: So if you’re taking pictures of crops, and just uploading random JPEGs of crops. Do they go back into your product and then have to tag the photos? And say, “Okay, if it looks like this, it’s soybeans. If it looks like this, it’s corn.”
24:25.62
Kirk Marple: So there’s something out-of-the-box machine learning. We do that. We’re using Azure Cognitive Services today. So we’ll do a first pass of tagging, but it’s [a] generalized model. So it’s going to find, like a building.
24:27.83
Matt Watson: And then that’s kind of [where] the machine learning part then happens.
24:45.21
Kirk Marple: But it’s not going to be like, this is a soybean. To get the knowledge if this is truly a soybean, you need some custom ML with audit. You can do that with auto-ML where the customer can tag their soybeans and say we’re delivering. Basically this is our big feature coming out next month; tag 20 or 25 soybeans [and] hit go. We do everything else, and then redeploy it back in their system. Then every other image that we see that has a soybean in it will get a soybean tag.
25:07.14
Matt Watson: Yeah.
25:14.97
Matt Watson: The challenge of that is actually more complicated too. I would imagine for the customers, because they’re gonna want to know like, “We just planted the soybeans. It’s grown 50 percent; it’s grown a hundred percent. It looks diseased. It looks like whatever.” And that’s where all the complexity gets into building all that in.
25:24.11
Kirk Marple: Yeah.
25:33.76
Matt Watson: But then at the end of this, you get something super amazing like a dashboard. You can build what percentage of our farm has grown this much and it’s healthy. It’s pretty cool stuff that you can do.
25:35.39
Kirk Marple: One.
25:41.87
Kirk Marple: Yep. And if you can bring it back to a baseline, and say, “We can figure out the date the image was taken. We know if it’s tagged with a soybean. We can track out those trends over time.” We do different things, but then the other goal is we do have plans to kind of map this back to the geospatial world. Or we can know that this is actually in this area of the world.
26:00.63
Matt Watson: Yeah, yeah, yeah.
26:17.42
Kirk Marple: And then we can track that specific area, have it evolve over time, and maybe even show like a 3D version of it. Then show heat maps of how the data set has evolved, so there’s a whole bunch of data visualization stuff. It’s a multi-year process. We’re taking a pretty big swing here that you got to start with ingestion. You got to start with just getting search and filter, and all that stuff working. We’re at the point now where we have the processing integrated, and we have the alert engine going in this quarter.
26:50.70
Matt Watson: Well.
26:53.19
Kirk Marple: So next month, that unlocks a ton of potential because now, you could be like, “Hey, hit me on Slack whenever I see a soybean or something. And then tell me or show me our daily report in email of all the soybeans that we captured in the last seven days.” We now have all the pieces in place to make that possible.
27:13.32
Matt Watson: What most people don’t realize is how complicated some of this stuff is. That doesn’t necessarily change the product like. So, for example, we talk about data ingestion, like you could spend forever trying to build a data ingestion pipeline.
27:22.84
Kirk Marple: Now.
27:30.87
Matt Watson: That performs well and scales and all that, but the end user doesn’t see that. They don’t appreciate that. It’s the hidden kind of [thing] when you’re building software. You spent months building this thing that nobody understands like you. You go to your investors like, “We got to go build an ingestion pipeline.” They’re like, “What are you talking about?”
27:34.72
Kirk Marple: Yeah, but now . . .
27:50.18
Matt Watson: But I’ve built one before so I appreciate it. You know, at my company, we ingest billions of data points. And when you start ingesting a large amount of data for a lot of clients, you have to worry about what if these people send us like . . . JPEGs right now.
27:50.84
Kirk Marple: Yeah.
28:07.65
Matt Watson: But somebody else just sent us 50,000 audio files to process. How do I ingest all this stuff? And oh, by the way, everybody wants everything in real time, right? So trying to handle the load and the performance where you store that data, and the scalability of all of it.
28:15.61
Kirk Marple: Right? Let’s do that.
28:26.00
Matt Watson: It’s not very hard if you’re dealing with one client and one project. But when you’re trying to build a multi-tenant SAAS product like yours, those are big challenges that people don’t appreciate.
28:27.71
Kirk Marple: You could put something together with an S3 bucket and a Lambda function. Maybe some kind of queue or something. I mean there are ways you can whip together a prototype of a pipeline. But then it’s, “Hey, okay. Well, do you want to have permissioning on that? Do you want to use azure and GCP and S3 [to] deal with all that?” And then you have that one model you’re running. But what if you want to run a model from another company too?
28:46.17
Matt Watson: Yeah.
28:54.70
Matt Watson: I [have] to feed it through machine learning sometimes. Sometimes I don’t.
29:02.98
Matt Watson: Yeah, yeah.
29:05.80
Kirk Marple: We’re trying to take all that heavy lifting away. It’s a classic platform as a service solution where it’s just refactoring hard stuff. You’re pushing [data] down and you’re standing on top of it. That’s really what we’re trying to do for customers. I mean, we’re sitting on top of Azure. Just the way that Azure sits on top of Intel.
29:09.77
Matt Watson: Yeah.
29:22.63
Matt Watson: You want your customers to log in and have these beautiful dashboards. But the real work is all done in the backside that nobody sees. That’s the reality of products like this, and products I’ve built in the past.
29:25.30
Kirk Marple: You know.
29:38.31
Matt Watson: 90% of all the hard work is in ingestion-aggregation. Like you send us a trillion data points, we have to make it so you can load that in like one second. When you load the dashboard, nobody appreciates all the work that goes in to make all this data fast to load.
29:53.78
Kirk Marple: It takes a little while. We’ve been going a year now, and we’re still pre-reve[nue]. We’re still getting there; to go to the market. But we kind of started. We thought we could sort of sell a Q4, just more of a search-view kind of interface on the data. But it’s like that actionable part was really the piece—the part people wanted. And they needed a bespoke ML like the out-of-the box Azure stuff. We thought we could be somewhat useful to see generic tags. But what we realized is [they don’t] need to say their stuff. Like they need to see a dam or an animal or this truck. It’s all in that last mile, so that’s why we’d really double down this quarter. The team busted butt to get all of this ton of new features. We redid the UX.
30:44.39
Kirk Marple: Next month is really our true launch and I think this will have the product that we can really get to market with. It was a bit too light on MVP in Q4. Everybody says, “Hey, try and sell the minimum.” But in this market, there’s a bucket of stuff you [need] to have before it really becomes viable. So we kind of learned that was the tough part. I mean, we learned where we were in October/November just wasn’t viable. But then again, we’d only been going for six or seven months, which isn’t a lot either.
31:16.45
Matt Watson: You have the kind of product where there’s a lot of different paths you could take, right? You could be like, “Okay, we’re a database. You send us the data, and we’ll show you dashboards about the data.” Odds are that they send you the data, but they need to do a lot of analysis on the data and stuff.
31:30.16
Kirk Marple: Yep.
31:35.97
Matt Watson: Improve the quality of the data first and then put it in your system. But then it’s like, “Hey, who does that? Do they do that? Or do you do that?” And that’s a professional service or consulting. Or like an integration project, right? You’re like okay.
31:41.80
Kirk Marple: Right?
31:50.16
Matt Watson: Going to do this three-month long project where you send us all these photos. And we’ll build machine learning. We’ll do all the tagging—that’s a lot of manual work. You’re no longer just the data storage. Now, you’re having to do all this other work on top of it. So, as a business, you almost have to decide which path you go down.
31:54.53
Kirk Marple: Yeah, yeah. Yeah, and I’ve done that professional services consulting as part of the product before [in the] previous company. I’m trying to stay away from it a little bit. Sure, it’s easy money at the start, but it’s hard to generalize the product.
32:08.59
Matt Watson: That makes sense.
32:24.30
Matt Watson: And it’s hard to scale.
32:26.60
Kirk Marple: To me, if you’re really made . . . you just scale into . . . I’m really trying to make this more like a Databricks-Snowflake kind of de facto standard for unstructured data. So you have to take a big swing and say, “We need to wait until we can do it for everybody and not just get sucked up with one or two projects.”
32:30.92
Matt Watson: Yeah.
32:45.22
Kirk Marple: Thankfully, investors have been really patient with it. We’re actually, as I said, raising now. But that’s why we kind of saw, like we didn’t raise a ton last year but we’ll do a seed extension now. Really just get more in the box, and then push hard in going to the market. Because you have to realize that when you’re selling a platform like this, it’s not a quick win. None of Databricks . . . It took a couple years for them to get sticky. Any of these big-swing products like this, you’re not going to see results in twelve months. I mean, it’s going to take that time.
33:18.15
Matt Watson: Well, that’s the thing. You have to find your niche, right? You figure [it] out. You build something in some sense. It could be this universal tool. It’s like, “Oh, we could take any unstructured data and we could do anything.”
33:23.84
Kirk Marple: Yep.
33:31.22
Matt Watson: But eventually, you may send around like, “Oh well, we really do really well with video files or audio files or images or drone images or whatever.” It’s like we built all this special stuff for machine learning that’s more computer-vision related. So it’s better for JPEGs or video files.
33:33.32
Kirk Marple: Right.
33:40.68
Kirk Marple: The good thing is it’s easy to pivot when you have a platform. If we wanted to build a vertical app on top of it, and that made us the same amount of money we’re going to make anyway, I don’t care.
33:47.81
Matt Watson: Sometimes you just have to find your way. You kind of figure out what your niche is.
34:00.68
Kirk Marple: We may find one or two kinds of vertical solutions that are our niche that blow up. I don’t know. But until then, we’re going to focus on building a wide platform. Because I really do think there are not just two or three verticals that we can approach. I think it’s dozens. The big thing for us is [that] I want to evolve it into like the knowledge hub of the organization. Because correlating all the data together is when it really gets interesting. And we pitched like, “You could have a tag in a document, in a spoken word on a Zoom call, and seen in an image. And we can connect that in the knowledge graph.” That’s unique. Nobody can do that, and be able to connect those entities that we pull out to external data. Like, “Here’s a record in a database or an SQL database.” That’s when you start to really truly spider out through your organization.
34:58.96
Kirk Marple: Pull and pivot on any of those angles, and start to be like, “Well show me all the recent drone images for this piece of manufacturing plant. But you can query it by the SAP record and get the drone data. And maybe build a 3D model [out of] that.” Just to be able to connect all those pieces is incredibly valuable. So, yes, we’re all Azure backend right now. It’s Azure Cosmos DB and then Cognitive Search. We kind of built a hybrid.
35:19.30
Matt Watson:Yeah, so you guys use a graph database.
35:36.12
Kirk Marple: Actually uses the SQL API—the Gremlin Api and Cognitive Search—in kind of a hybrid way.
35:38.94
Matt Watson: For those who aren’t aware, a graph database is a newer type of database that relates entities to each other. As you said earlier, think about feeding a bunch of text, PDFs and even audio files. And they find Matt Watson in all of them, right?
35:47.62
Kirk Marple: So . . .
35:58.81
Matt Watson: You’re able to create those relationships. [For example], this file links to this file, which links [to this file]. You build a graph, like a tree almost, of how all the things relate to each other. Which is pretty cool stuff.
36:08.27
Kirk Marple: And the great part about it is you can keep enriching that over time. Once you have the data in there, it’s not static. You could run more and more algorithms on it, and then you can even infer clusters of tags. You [will be] seeing we could cluster in time; we can cluster geospatially. [We can] cluster in this kind of tagging concept. I can’t wait until we can get to a point where we can start running algorithms on the graph as a whole to do similarity search. And say, “Hey, that’s weird. 20 percent of your data over here doesn’t have all these common properties. But it doesn’t have this tag.”
36:46.34
Kirk Marple: So do you want to autotag with this tag? Because maybe there’s all this kind of quality data lineage questions.
36:51.26
Matt Watson: Part of your challenge comes to you . . . are you creating a product that’s generic for everybody? Or are you creating an industry-specific solution? As you mentioned, there was a company in Kansas City called Claim Kit, which I think sold recently. I’m pretty sure it’s related to all this.
37:10.85
Matt Watson: They were scanning PDF files of insurance claims or legal documents. They were scanning legal documents to understand all the legal language that was in the documents. Like what did we promise? What are we on the hook for? [They] have to scan all these PDFs and all these legal documents.
37:12.67
Kirk Marple: Right.
37:19.80
Kirk Marple: Um, yeah.
37:29.27
Matt Watson: And then basically get the metadata out of them. To know like, “Hey, we have thousands of these legal agreements that we agreed to. Like you’re an insurance company [with] all these policies. And then building a database on top of that to know all the policies that exist, what did we promise, and what were the terms.
37:29.38
Kirk Marple: You have.
37:40.13
Kirk Marple: Um, yeah, yeah.
37:46.17
Matt Watson: And that’s an example of an industry-specific, similar kind of solution.
37:46.79
Kirk Marple: I think there’s overlap where anytime you’re creating structure from sort of the quote unstructured data. I don’t think we could probably never support that full solution. But we could probably carve away 68% of the hard work.
37:55.77
Matt Watson: Yeah.
38:02.40
Matt Watson: Yeah.
38:04.72
Kirk Marple: Our goal would be . . . We’ll handle the data storage, the search, the query, [and] scale all those things. If you want to add your own APIs or any kind of computer, we can trigger you. Like we have a new PDF and we’ve done a first pass of analysis on it. We can say, “Oh, yeah. Since it’s an insurance form, you can do the bespoke part yourself.” So our goal, our hope, is [that] a company would start with us . . . building vertical products.
38:32.16
Matt Watson: Yeah, yeah.
38:43.12
Kirk Marple: They would only have to build that last twenty percent. And not have to build all this junk. Nobody writes their own database. Nobody writes their own CloudStack. So we’re just trying to be another layer in that stack that essentially . . . You could go into the Salesforce ecosystem and write an app in the Salesforce ecosystem.
38:51.12
Matt Watson: Right.
39:01.81
Kirk Marple: You get all this other power and all this data modeling and all that kind of stuff. Essentially the same thing we’re trying to do here.
39:03.19
Matt Watson: Yeah, yeah. A good example of that. I think it’s important for our listeners to understand. [That] as you’re building a platform, you’re not necessarily building the product. It’s building a database [but] somebody has got to do something with it.
39:22.62
Matt Watson: Back to the Claim Kit example. They built a product where they could have used your platform. It might have saved them a lot of headaches. Like, “We use their platform to get all this data. [We] ingest it like that was the big problem. Tag it and find all the information we need. And then we were able to build like business intelligence and like the final product on top of the platform.”
39:39.27
Kirk Marple: Yep.
39:42.57
Kirk Marple: That’s exactly what I mean. We talked to a power company, which would be crazy to think a power company is doing computer vision. But they are and they’re doing it to identify the metal tags on wooden power poles. Basically, they were writing their own computer vision to do text-to-text extraction from that.
39:58.36
Matt Watson: New here.
40:01.15
Kirk Marple: But when you talk to them, they literally had to build soup to nuts. Like a queuing system with Kafka, a data ingestion system, a database. And they built [it] from scratch, so they had to waste time, data, [and] team effort. But if we had been ready, and I met them probably almost two years ago at this point.
40:06.31
Matt Watson: Um, yeah.
40:20.76
Kirk Marple: We could say, “Look here. Just start with us. Here’s a bucket used provisioned for your customers. Dump data in there. Configure it to run this kind of ML. And then figure out what’s left. What’s the bespoke stuff you guys need to do on top of it?” I see them as a perfect kind of . . .
40:29.18
Matt Watson: Yep.
40:39.94
Matt Watson: You’re right.
40:40.22
Kirk Marple: So we should provide downstream. Build on us as a platform—as a service—and it cuts away so much heavy lift.
40:45.74
Matt Watson: As a software developer, part of our problem is we see something we like . . . Oh, we could build that. I can do that this week, and I’ll put it in an S3 bucket. I’ll use AWS Lamden. And then I’ll get the metadata and I’ll show the metadata and MySQL. Then we’ll just query MySQL, like I don’t need this thing, right?
40:51.69
Kirk Marple: Yeah.
40:59.56
Kirk Marple: That. You.
41:04.77
Matt Watson: No doubt there’s people that do that, right? But then, you have people that also are able to realize that “I don’t want to maintain that thing” and “it’s not going to scale” and assign these use cases. It doesn’t do all this machine learning. It doesn’t have all these dashboards. It doesn’t have all these alerts. it doesn’t it doesn’t have all this cool stuff.
41:10.54
Kirk Marple: Right.
41:24.80
Matt Watson: That’s part of your challenge too, I’m sure. With some of the people you pitch. They’re like, “Well, I could just build that.” You’re like, “Oh, wait. At some point in time, you guys have enough features and products to build out.” They’re like, “No, we’re not going to build that. It doesn’t make any sense. It has all these added benefits that my developer’s not going to build over the weekend.”
41:32.65
Kirk Marple: Right? And that’s what we’re just getting to. That tipping point, now, where it’s like I can show off. We’re trying to really follow the kind of no-code plan . . . you don’t have to be a developer.
41:47.70
Matt Watson: Um, yeah.
41:52.35
Kirk Marple: You just go in. It’s like Zapier or something like that. You just connect up a couple things and boom! Just drop the data and it works. It’s really interesting to see how this has evolved. And I think it does take some learning for people to catch up because I think it was a little bit early for the market to really understand it.
41:55.62
Matt Watson: Right.
42:11.48
Kirk Marple: But all the signals are there. You’re hearing all the podcasts [that] by 2025, the amount of data that’s getting pumped through on structured data. Everybody is overlooking this area because it’s all about the structured world right now. That’s why I’m just trying to beat it.
42:29.57
Matt Watson: Yeah.
42:30.30
Kirk Marple: I think it’s just getting that market acceptance is key.
42:34.83
Matt Watson: There’s been a lot of players that have been doing unstructured data. But it’s mostly been on text data like Splunk. But it’s totally different when you’re dealing with media files. Like you mentioned earlier, I’m sure the government does this.
42:41.63
Kirk Marple: Yeah, um, yeah.
42:53.75
Matt Watson: Transcribing audio files and figuring out they’re talking about Vladimir Putin. I’m guaranteed that you know the government does this kind of stuff. And there are tools that do it. But the question is—are there tools that a normal business would do? Has the market got to a place where people need more of that kind of thing? And you can provide a platform to do it.
43:05.60
Kirk Marple: I pray.
43:10.73
Kirk Marple: If you look at Otter.ai or one [that] transcribes your Zoom calls. Or what’s the one [that] does this for sales calls? Essentially, they would have to build a lot of this similar ecosystem.
43:12.56
Matt Watson: Right. Yeah there’s a lot of them.
43:29.69
Kirk Marple: Running data ingestion [and] searchability . . . Running some sort of analysis. And I’m sure every one of these companies has something that looks kind of what we built.
43:37.24
Matt Watson: Right.
43:48.72
Kirk Marple: Here’s another layer of the stack that you can just build on. Hopefully, that lets people innovate more where they’re not wasting the first year of their company. But it’s like I lived in the on-prem world. I was setting up servers at the colo; I don’t want to have to do that again.
43:54.88
Matt Watson: Um, right, It’s a lot of plumbing.
44:06.47
Kirk Marple: So we love living on top of Azure. Just the way we hope companies will love living on top of us.
44:11.22
Matt Watson: Yep, well, very cool man. Once again, a big thank you to today’s sponsor, Universal Registered Agents. Set up your new business and maintain all aspects of your business compliance. Their goal is to make your job easier, so you can focus on what you do best—running your business.
44:26.62
Matt Watson: Connect with them by visiting the link in our show notes. It’s the same pitch for you. It’s like, “You don’t want to deal with building your own ingestion pipeline and how you process all these files. Let us do it and build intelligence on top of it.” Like it’s beautiful.
44:38.46
Kirk Marple: It’s really like starting the company. There are so many great tools out there, but just for banking, credit cards, and all this for the company. Now, it’s almost like the business has got refactored. You could basically have your business stood up with . . .
44:51.19
Matt Watson: Um, yeah.
44:55.68
Kirk Marple: Best of class ecosystem for the business in a day. And it’s been really fun to see.
45:00.18
Matt Watson: Very cool. As we wrap up the show here, any advice to business owners about their data.
45:07.35
Kirk Marple: Don’t hold back in terms of what you think it’s capable of. I think [that] is the biggest thing I’ve sort of seen. There’s almost a glass ceiling of people who don’t realize they can do as much with their data as they can. They may think that “I can just search by file name or in a folder.” There are new ways out there to get more value out of your data and analyze it better. So try to learn, educate yourselves with podcasts and blog posts, and stuff like that. There’s a lot more out there than probably people realize is possible.
45:41.18
Matt Watson: All right, very cool, man. Well, thank you so much for being on the show today and nerding out with me a little bit about data. I’m also a data nerd. Very cool stuff, and I hope everybody learned some things today. Thank you sir.
45:53.88
Kirk Marple: Thanks so much. I appreciate being here.