 Hello, and welcome to Data Diversity Talks, a podcast where we discuss with industry leaders and experts how they have built their careers around data. I'm your host, Shannon Kemp, and today we're talking to Ryan Welch, the founder and CEO at Kendi. Hello and welcome, my name is Shannon Kemp, and I'm the Chief Digital Officer at Data Diversity, and this is my career in data, a Data Diversity Talks podcast dedicated to learning from those who have careers in data management to understand how they got there and to be talking with people who help make those careers a little easier. To keep up to date in the latest data management education, go to dataversed.net forward slash subscribe. And today we are joined by Ryan Welch, the founder and CEO at Kendi. And normally this is where a podcast host would read a short bio of the guests, but in this podcast your bio is what we are here to talk about. Ryan, hello and welcome. Hello, thank you for having me. So tell me, founder and CEO at Kendi, so what is Kendi and what do you do? Yeah, we're a natural language processing company, so we help enterprises get value out of their unstructured text data. Specifically, if you kind of think about NLP, it has five, now probably six major categories if you were to like look in a undergrad textbook. One of those categories is natural language search. And so we help enterprises primarily with natural language search. They can interact with our software via the search bar or we've been seeing more and more chat interfaces. So it's interesting times. Very cool. So who are typical users of your product? Yeah, the users can be internal or external folks. And so if you're an external customer coming to a website wanting to learn about a company's product and services, you can use Kendi natural language search to ask questions of the unstructured text data on the site and actually be pointed to the answer on the site. So it's super easy to use. And then also internally employees will use the service for searching policies and procedures. And so it's all types of other stuff. So it's really anywhere where there's unstructured text that people are trying to find the answer. So tell me, so when you were very young, is this what you wanted to be when you grew up? Like, I'm gonna grow up and be a founder and CEO of a natural language company. No, no, not at all. Honestly, I'm kind of jealous of people that have a really early understanding of what they want to be. Some people that I grew up with or knew later on, they're like, I always knew I was going to be a lawyer or a doctor. And they've been working on it since they were like nine. And I got super jealous of those people. That was not me. You know, I have so many, so many interests that you just kind of hop around and you just find things that are interesting. But I would say like the one thing that was always of interest to me and maybe it's all those interests is understanding things. And I feel like people who work in data are just curious. And their natural inclination is to then go, all right, well, let me go read something. Let me go download a bunch of data in a spreadsheet and see if I can understand it. And so, no, I didn't know that I was gonna be a founder and CEO of a software startup. But I was always curious about learning new things. And I think that just then came to, you know, analyzing data and looking at data. And here I am. I can totally relate to that. I suffered from that jealousy myself. Yeah, it was like the other, like last year, I was thinking about buying a home. And like, it was funny, which was the worst time to buy a home, particularly in the Bay Area. Everything was like 100% more than the year before. But like immediately downloaded a bunch of data from, you know, I think was Zillow or Redfin or one. And I have like basically the years, you know, two or three years worth of housing data with square footage and I'm applying, you know, statistical models and predicting home prices. And, you know, and this is in the hour that I have before bed. So, yeah. Well, I love it. Well, so how then did you get from to where you are? Like, where did you start out? What did you study? Yeah, my undergrad degree is actually in anthropology with a minor in math. And then I went and did a graduate degree in applied math, which is basically just, or at least in this instance, economics. Then I went and did a economics and like quantitative finance. And then I went and did a MBA at the University of Notre Dame. And when I got out from the University of Notre Dame, I actually started working for a group of folks in the DC area, helping them commercialize technologies coming out of national laboratories. And so there's a lot of like really cool tech that the national labs create, whether it was like low orbit satellites, defensive cyber capabilities, quantum cryptography. And so it's like all this really cool technology. And I was like, man, you know, maybe I could start a business around something in the deep tech space, which then kind of led me to think about, you know, where is the world going? And, you know, in 2012 with deep learning making some pretty significant improvements, I kind of focused on the natural language space. And so that's how I kind of like evolved to get here. Just kind of having a major macro thesis on kind of where the world was going and then taking a bet on myself to build something to solve some problems. Well, that's very impressive. And that certainly took some courage. So what made you finally pull the trigger and say, okay, this is what I'm doing. This is what I'm gonna develop. What made you take that leap? Yeah, the biggest, well, I'll get there, but someone wrote me a pretty sizable check and said, come back when you have a business, which was hilarious. And so I had a macro thesis that, you know, and this was in 2013, 2014, where I said by 2025 all knowledge workers would need some form of AI or machine learning and able to workflow. And specifically around unstructured text data, because I felt like we were the bottleneck in the modern production process. And when I say we, I mean people, because we can store a bunch of data, we can move a bunch of data back and forth. We can even process a bunch of data provided that it's structured and computers can actually process it. But on the unstructured data side, like as soon as it hits our eyes, you and I can only read at a speed that is similar to what our parents and grandparents can read at. And so we're still reading at a page a minute, you know? And so it's kind of like, oh, that's interesting. So like in a modern knowledge economy, we're effectively the bottleneck in this modern production process. And so it was like, could you create technologies to kind of amplify the productivity of people to kind of increase that output around these knowledge intensive tasks, which underlying those tasks was unstructured text data. And so just kind of having that macro thesis, I asked around for some technical co-founders and a gentleman that I was working with here an early investor in Kendi. I got back to his office and he pulled out his checkbook and he wrote me a $100,000 check and said, come back when you have a business. I was like, yes, I need to start a business now. That's amazing. That's quite the story. That's very nice. Yeah. He's an incredible, incredible angel investor in the DC area. His name is Jim Huntsall. Say his name because he's a venture capitalist over at Laugh Rock Ventures here in DC, even though Kendi, we're based in the Bay Area. So I'll give a shout out to Jim. There's probably several dozen people like me that Jim had done that too. He's also a professor over at Georgetown who teaches entrepreneurship and venture capital over there. So just an incredible person. And he probably did, you know, single-handedly built the DC tech ecosystem. I love it. I love stories like that. I mean, it's just, relationships are so important and how we move forward. Visit dataversity.net and expand your knowledge with thousands of articles and blogs written by industry experts plus free live and on-demand webinars covering the complete data management spectrum. While you're there, subscribe to the weekly newsletter so you'll never miss a beat. So tell me, I mean, with all this experience and in your diving into this, so what is your definition of data? Yeah, my definition is, and this is a bit of a loosey-goosey, that definition is any information that is annoying to comprehend manually. And so it's like, and so it's like, you know, if you have two cells in an Excel spreadsheet where it says like 2021 revenue, 2022 revenue, that's probably not data because you can look at it and be like, oh, 2022 revenue was higher or lower than 2021. But if you have data going back 50 years or information going back 50 years and you wanna know that, hey, how much did it go up compared to 50 years ago? And you're now doing changes in that. Like you've now crossed over into the realm of, that's data, even though it's pretty small data. It's not, you know, terabytes and petabytes of data, but it's large enough to be annoying to just quickly comprehend manually. I love that definition. That's fantastic. And with these predictive models that you've had and this analysis that you've done and your thesis, you know, do you see the importance of data management and the number of jobs working with data increasing or decreasing over the next 10 years and why? Yeah. My marketing team's gonna be a little upset with me because so we're gonna make a, I'll say it here first, but I actually, you know, we'll come out with the kind of predictions thing for 2023, but I actually think we're past peak data scientists, which is a bit of a hot take. And the reason why is, and when I think of data scientists, it's not necessarily someone that can like just comprehend and use statistical tools to understand data, but like there's been this, you know, for the last several years, like automation of certain data science techniques. So you can think about like auto ML, there's now like auto NLP. And so like, do you really need to like build algorithms? Do you need to do any programming to do complex statistical analysis anymore? Probably not, right? I mean, you still need to understand like, what are the right techniques to it, to apply it to data and, you know, what are the right techniques in actually understanding any biases in the underlying data or the technique that you're using. So there's still like that type of data science, all called like job that's gonna be available, but like this previous notion of like, hey, I'm gonna be a quant finance person or I'm gonna be a data scientist like writing my own code and doing my own analysis. Like I think we're past, I think we're past that. And I think we're past it because a lot of vendors have created a lot of really awesome capabilities to get around that. I think it's a major bottleneck in the deployment and the success of AI and machine learning in the enterprise. And so I think a lot of people have gotten around like, hey, let's just skip the data science team and go right to developers and right to line of business or can we make it easier for the data science team to build models and pass those off to the developers and line of business people who actually put them into production. So candidly, I think we're past peak data scientist and yeah. Do you see it shifting then to need more analysts or? Yeah. Yeah, I kind of think about like data sciences as evolving to be kind of like a business analyst. Like, if you think about a business analyst, like they're not a coder, they're not gonna write product code that is world-class awesome code, but hey, they're technical, right? And they understand how to do SQL queries and they know some things, but most importantly, they know the business problem and they know when like management comes to them and says, hey, can you show me how many red shoes we sold in Kentucky over the last 50 years, like they'll come back with a visualization and show that manager and they'll be able to manipulate the data and kind of do those queries and get that information back into a visualization capability. And so I think like kind of, yeah, data scientist is gonna look more like a business analyst in the future than what was previously like this, like hardcore data science. You know, like I remember in graduate school using, what was the, what was, it was like SAS, SAS, the kind of more and more advanced statistical packages, like, you know, I remember writing, you know, or programming to do advanced kind of statistical analysis on economic data and financial data, but this was in graduate school, so that'd be like 2003, 2004, 2005. You don't need to do that anymore. You know, I upload housing prices to a data robot or an H2O and I hit go, and it'll pick out the best model for me. So that's where I think we're going. That's very cool. And then also on that fringe, then where do you see like data architects and modelers fitting in there? You know, I mean, to set up those really cool tools and to ensure they stand up and maintain that quality and so forth, yeah. Still critical, still critical folks. I mean, I think that's still a hard problem for a lot of enterprises is like pulling all this data together, getting the right kind of like, just working environment for people to get value out of the data that an enterprise has. So I think all that's still gonna maintain some strength there, but it's like, I don't know, you know, maybe it's just me, but I kind of like, you know, envision like this data scientist as like the sexy thing where it's like, I'm gonna go do this awesome thing and don't bother me for a month and I'm gonna give you insights that you've never seen before and your mind's gonna be blown and your hair's gonna be blown back and it's gonna be incredible. And so I feel like that is kind of like, we're a bit past that, but in a good way, right? Where it's kind of like, you know, do I really want to go back to writing like SAS programming for advanced to disclose? Absolutely not, right? We want to make it super easy. So maybe data scientists can focus on, you know, higher value things than some of the lower value stuff that was kind of backlog. With a robust catalog of courses offered on demand and industry leading live online sessions throughout the year, the Dataversity Training Center is your launch pad for career success. Browse the complete catalog at training.dataversity.net and use code DBTOX for 20% off your purchase. Very cool. Definitely a bigger ROI in that. Absolutely. And I think there's a bigger ROI in text data. And so I kind of have this, you know, if I were to compare things to like financial markets, I would say like the edge that people can get now is all in text data and all in unstructured text. So I feel like all the structured numerical data has been manipulated and modeled and, you know, combed over because it was easy to look at and easy to analyze and easy to apply statistical models to. And I think like the real edge in anything is getting those insights out of unstructured data, which is really where the world's been going for the last, you know, 10 years now with, you know, whether it was computer vision or, you know, video analysis. And now it's all the incredible stuff that's coming out in natural language processing. Like that is an incredibly hot field that I think is really compelling and super interesting. Indeed. So then what advice would you give to people looking to get into a career in some aspect of data management, whether it be an analyst or a data scientist or? Yeah, do it. Like, you know, when I say that, like, you know, I don't know, you just got to go do it. I mean, there's things that you need to learn. So when you're doing an advanced kind of more, you know, more dense statistical analysis, you know, you don't, you want to understand the techniques that you're using. So one of the, like, potential downfalls of, like, all these automated tools is that they're all automated and you actually don't know what's going on under the hood. Whereas, like, previously, like I remember my parents telling me about math. They're like, I know math because, you know, we had to do it, you know, really manually. And I'm doing it on like a TI-83 calculator and plotting charts and stuff like that. And so one of the challenges is you start to then, like, not really understand the underlying fundamentals. So, like, if you're really interested, I would, like, recommend folks to, like, really get down to the nitty gritty of, like, understanding all the different techniques. But then specifically understanding kind of what types of data you want to look at. Because it's funny, like, you know, when you talk to someone in the AI space and, you know, friends will say, oh, Ryan, you work in the AI space and this person over here works in the AI space. You should get together and I'll talk to the person and they're in computer vision and I'm in NLP and we have nothing to talk about. We have nothing to talk about. You know, maybe we can, you know, bond a little bit over transformer models and stuff like that, which is kind of more common these days for this kind of analysis. But, you know, you're really going to have to focus your tool set on the type of data that you're going to want to look at. And so, you know, maybe there's a type of data that really appeals to people. And if so, you know, focus on that, learn the tools that are best for that and then just get after it. Yeah, so, you know, with data scientists and analysts, those are a couple of, you know, degrees that people can major in, you know, but what's not being taught in school? What's missing from that curriculum for the practical? Is it just learning the tools and the technology and how do you keep up to date on the latest tech? Yeah. Well, I mean, tons of papers. So I'm like still like just like in the research. So I love like just reading all of the research articles so you can like follow folks on Twitter and LinkedIn. And there's a lot of people that are posting just awesome stuff that's consistently coming out. And, you know, like in our Slack channel, I'm like consistently posting like the new leading edge AI papers from archive. And it's just like, my team's like, all right, we get it. Like I'm going to stop posting all this new stuff. So I'm constantly, I'm a very voracious reader. So I'm reading all that information. So that's how I generally stay up to date on things. And then as far as like the different tools, like, you know, it's really just, I think about like just, I mean, just doing it. Like if you're a curious person and you're getting into data science, then you're just going to be doing this stuff I would imagine just in your spare time. So just continue to do that. You'll get really good at it. And then someone's going to pick you up as a, for a job and you'll be doing it professionally. And now you'll be doing what you love professionally. I love it. So it's been kind of a theme through the, through the conversation is be curious, like curiosity has moved you forward. It's a way to keep moving forward in anything. And it's a great message. Yeah. And like I said in the beginning, like I think people that work in data are just naturally curious. And then you're trying to pull down a data set to analyze it. And also like there, I mean, except for maybe some, some startup founders, most people in data science are a bit humble too, probably. Like in the sense that like, I don't know the answer to the question. So let me analyze it. So like, oh, that's a curious question. Let me go try to pull 10,000 articles and see what the answer is. Like people would normally wouldn't say that or let me go see 50 years worth of housing data and see what the, you know, if I can predict 2023 housing prices, like people normally wouldn't do that because they'd probably normally wouldn't, you know, question themselves going like, hey, do I actually know the answer to this question? And, you know, a lot of times you'll say, no, I don't actually understand this. So let me actually use the tools that I have that I've learned, you know, from undergrad or a graduate degree or some accreditation course to actually then pull that data. I will also say probably a critical thing has probably become more and more important is really understanding like the business case for it. I think one of the biggest challenges today in the enterprise that we see is just not a clear connection between the folks that are kind of doing the more advanced modeling and like the business outcome and what the line of businesses is trying to do and try to do. And so like, if you're in a data science course or you're a statistics major or something like that, if you can take business courses and just really understand like, hey, what's the business actually trying to do? Like what is marketing trying to do with this? Oh, marketing is trying to target specific people so that they can get leads to fill the top of the funnel which then goes to the sales people who can then close deals, you can then generate revenue. Okay, now I know what I'm doing. You know, what business KPIs I'm driving. So if there are just kind of basic business courses that you can take to really tie all that together, I think you'll have a leg up on, you know, most people in the data science space. That's very sage advice. Is there a couple of resources that you are your go to that your favorites for your articles and research? That's a great question. I actually can't think of any top of mine because I just pull from like all the social networks. And so, you know, folks that just post stuff so again, whether it's Twitter or LinkedIn or, you know, Google alerts and stuff, stuff like that. I just pull all that and I can't say I remember where the actual URL is too. Your favorite sources through networking. Yeah, exactly. I love it. Well, Ryan, this is amazing. Thank you so much. And I'll be remiss if I don't ask, you know, if people want to find out more about Kendi where should they go? Oh, yeah, they should definitely go to Kendi.com, www.kendi.com. I feel like you have to say that. And then, you know, we have tons of white papers there. You know, folks can tons of content on the site. People can reach out, happy to set up demos and just talk to anyone and kind of share more about what it is that we do. I think the natural language search space is just super compelling and interesting space right now. A lot of companies are popping up large and small. A lot of money from the venture capital spaces is going into natural language search. So, yeah, they should definitely, definitely check us out and, you know, always happy to chat with people. I love it. Well, thank you so much. Anything else you want to add before we wrap it up? No, I think that's everything. Hopefully it was, you know, it provided some good advice to the folks out there. Oh, very sage advice. So really appreciate it. Thank you so much for taking the time to chat with us today. And for all of our listeners out there, if you'd like to keep up to date on the latest podcasts and the latest in data management education, you may go to dataversity.net or slash subscribe. Until next time. Thank you for listening to Dataversity Talks brought to you by Dataversity. Subscribe to our newsletter for podcast updates and information about our free educational articles, blogs, and webinars at dataversity.net, forward slash subscribe.