Navigating the Data Frontier: A Conversation with Databricks Chief Evangelist Ari Kaplan

0:00Welcome to the AI Chat Podcast. I'm your host Jaden Shafer today on the podcast we have the pleasure of being joined by Ari Kaplan who is a leading influencer in analytics artificial intelligence and data science. With a career that has touched everything from Major League Baseball to Fortune 500 companies, he's known as the real Moneyball guy. He has spearheaded transformative AI and analytics projects and is the co-author of five bestselling books in these fields with roles that are ranging from Databricks head of evangelism to leading the Chicago Club's analytics department. Ari's diverse experience positions him as a pivotal voice in the AI community.

0:37Welcome to the show today, Ari. Hey, thanks. Great to be here. I've loved listening to a lot of your podcasts and appreciate everything you do. Super exciting, very eclectic group of people and always inspiring. Yes. Yeah. Very eclectic. Super excited to have you on the show. As I said, something I'd love to kind of kick this off with is asking you a little bit about your background and your journey, you know, in tech specifically. Did you always know that you are going to kind of be in tech? Is there always something you're interested in AI now today? Or is this something that you kind of discovered as you went through?

1:10Tell us a little bit about, walk us through your kind of background and journey. Sure. No, I'm a naturally curious type of person. So I love exploring new things and really in technology and innovation. So I find like through my journey, every like three or four years, there's some like major shift. Generative AI right now, more traditional AI, you know, four, five years ago, Lakehouse technology, database, you know, and you keep rewinding. So it's always really been in technology and I think kind of drives with the human fascination of taking the real world and trying to quantify digitize it, whether it's old school chemistry from 500 years ago to the physics.

1:54How can you take something in the real world and automate it and repeat it through technology? Yes, very cool. So that's incredible. I think curiosity has led a lot of us into this space. So much is changing. So much is evolving. So that's really, really cool. Something I want to double click on from your past that seemed really interesting to me, for those that have seen the movie Moneyball, they'll know what you're talking about here. But tell us a little bit about your role, how you got into Major League Baseball, and how your role in analytics kind of played in that whole thing. And maybe explain for people that haven't seen the movie?

2:33Sure. No, it's been a real honor that most of my career has been around sports analytics. And, you know, when I, the journey has been, you know, pretty unique. When I started, I was one of the first four people that have been known to work in some capacity is what you might call like a data engineer or data analyst. Okay. So I was just a student at Caltech. If you've the Big Bang Theory. That's kind of the school. I resonate. It feels like it was based on my life a little bit. Characters were spot on in how they behaved. But yeah, I was an undergraduate and had this summer undergraduate research fellowship to just show that there are statistics that were out there that are better than were commonly used and was able to explain that and communicate that in simple terms.

3:23And Fred Clare, who was the general manager of the Dodgers, had heard me starting to get into the news media. It is Hollywood, LA area after all, and said he thought it might help him identify or trade or work better with his team. So he invited me up and did some work immediately with some of future Hall of Fame players, Oral Hirschhiser, or Kirk Gibson. So that was like way before Moneyball, but always had to keep reinventing from coming up with better ways to analyze players. And mind you, these are like significant decisions. If you get one player different, it's playoffs or not. But then merging into scouting technology, how do you quantify what scouts are saying so that it's like everyone's talking on the same scale and people like who are non-technical um maybe we can talk about uh you know that democratization aspect but back then it was the democratization of being able to query data uh without having to know sequel or or information yeah and then well moneyball was like a great movie uh my opinion and book about how this like really changed the whole industry how they were first to move in some regards, and that helped get them good players with lower cost.

4:47And now, fast forward, the movie's been an inspiration to a generation. It's been over 20 years, and every team went from zero to one data scientist analyst to 30, 40, 50 for each team, and now went from an industry of about four people to tens of thousands of people if you include like vendors and the people that work for teams all around the world so that that was like the movie uh changed my life uh and then it the the the great thing actually happened after the movie after every team implemented analytics it changed the game so much that they just last year had to change the rules to like limit defense shifting since it got quote unquote boring.

5:34It worked too well. That's so funny. Yeah. I mean, I feel like with any kind of arbitrage like that, whoever gets it at the very beginning. So kudos to yourself for kind of being on the cutting edge of that sees the major benefit. Eventually, when everyone adopts the same strategy, it's a little trickier, but that's awesome. And yeah, I think it really kind of goes to show the importance of focusing on data with these players, obviously beyond just, you know, their skills or what you think they can add to the team, really looking at taking analytical approach. It's awesome in sports, but that also applies to so many different areas.

6:07And so I think, you know, that whole movie and movement kind of helped raise awareness for that, I think, for a lot of different industries as well. So very, very exciting. So currently, you are the head of evangelism at Databricks. Tell us a little bit, I would be curious, like, how did you come into that? How did you meet Databricks? What's your story with them? Yeah, great story. And if you haven't heard the title evangelist, it's kind of a newish, newer trend that tech companies are doing. It's not like preaching religion, specifically, but it's, you know, like a brand ambassador, somebody who can, like going on here with you, Jaden, people who've never heard of us explain it, or explain it in simple terms, sometimes things get a bit technical.

6:54So kind of like, what do you do and why? And I had that role a couple times before my prior job. I created and led evangelism at Data Robot, which is auto machine learning. And then I was the head of the user community worldwide for Oracle, a big database company that when we acquired MySQL, Java, people know of, PeopleSoft. But yeah, with Databricks, a quick growing company. I'll explain who we are in a minute. But, you know, after a certain size, when you're certain growth, you know, you need to expand and build out your brand. And so a role came up and rewind all the way back to that Oracle mention.

7:38But Rick Schultz, who's our chief marketing officer, he and I go back to our Oracle days. So there's that good personality, you know, that relationship, which is so important. And so, yeah, that came about and now it's a great opportunity. Databricks, if you haven't heard of us, we created what's called the Lakehouse technology. And that joins the old school, like structured database data warehousing, like Oracle, SQL Server of the world, and unstructured data that they call Data Lake. So machine learning, video, audio, you know, PDF things. You might have heard of the data swamp. It's like ungoverned, just like pop data into files.

8:27Yeah. So the lake house solves the problem of having one unified environment where you have all that data in one place. And it's based on open source technology. Our founders actually created Apache Spark, MLflow, Delta Lake, which gets hundreds of millions of downloads a year blows me away. Crazy. But because it's, yeah, it's incredible. It performs typically, you know, many times faster. You know, everyone is different, but multiple times faster than traditional Data Lake data warehouse. And it's oftentimes 10, 20, 50, 80 % less expensive. And it's simpler. So usually you have to pick one of the three.

9:12less expensive, faster, or simpler, but you get the benefit of all three. So that's in a nutshell, Databricks, you know, we did make news a couple of weeks ago when we raised an additional round, which funding, which makes us the third largest private tech company in the world behind SpaceX and Chime. So incredible. Yeah. And congratulations. It's, you know, and to your, to your and the team's credit, looking at a lot of the moves Databricks is doing, I've been super, super impressed. And I think it was just earlier this year, right there, a big merger happened between Databricks and Mosaic ML.

9:49So I see Databricks really doing some incredible moves in the space. What are some of the things that you're most excited about that Databricks is currently working on or implementing? Yeah, there's so many exciting things. And our core lake house technology, what we've been doing up to this date, it's now 1.5 billion a year in revenue, and it's focused on data and AI. Everything from the data engineering, how do you get these pipelines and these flows to the machine learning, to the ML ops, to operationalizing it. And so that's all exciting, what we're doing at the core. And what we have now in building, the whole marketplace is exciting, but what we call unity catalog, which is everything from the governance.

10:37How do you do audit trails? How do you understand what version of your data is accurate? What version of your data is from five years ago? You want to understand that to the transparency. So understanding that we taught, everyone's talking about LLMs and we We don't know if it hallucinates, but having that governance to understand where your data comes from, maybe have it privatized so things that are private to your company don't get out in the world, that is getting more and more important. So that's our Unity catalog. And then you had mentioned Mosaic ML. That was the large acquisition just a couple of months ago.

11:15And building large language models is like all the attention these days. I think it'll get back a little bit where we'll be large language models and more traditional predictive modeling and classification. but that's super exciting. The use cases that are coming out and that are going to come out that we haven't even thought of is super exciting. And then looking even more forward, Databricks, other companies, we are looking to leverage AI within our products. So as you're a coder, you're writing Python or pick your language, it helps. It used to just be autofill the name of the table you're looking up right ai can like help suggest comments and do quality assurance synthetic data generation um so it's like a code assist a co-pilot as you code which is really exciting and what we're able to do now is great and one year from now two years like the the world of software development is going to be uh like advanced well well ahead of what it is now Yeah, I see a lot of changes and shifts.

12:30And it's kind of interesting because I feel like when you look at a lot of different players in the space, Databricks is definitely one that is going to be a really big winner in this whole AI space. And for good reason. They're doing some really cool things, making some cool acquisitions. I believe you just raised$500 million at a$43 billion valuation. So congratulations to the whole team on that. really exciting to see, you know, the market's confidence in essentially your strategy and your plan. Something I'd be curious to ask you about is, you know, what do you think sets Databricks apart from perhaps competitors or other people that may try to do some of the similar things?

13:05What do you think is, you know, the secret sauce and some of the uniqueness of Databricks? Yeah, yeah. You know, great question. And, you know, we've had a lot of success creating the Lakehouse. So, you know, there are, and there will be other companies coming into the space. You know, all different things. Number one is we were built like the whole company, we created Lakehouse and the foundation was open source. So as an effect, like we're not taking existing software and trying to tweak it and back fit it. So as a result, you know, we have companies running their own benchmarks against, you know, anything out there and And benchmarks, by and large, are way, way faster, way, way less expensive.

13:50So people, especially in this economic environment, where you have to kind of watch your bottom line migrating. The other thing is the scalability. So like data bases. I love Oracle. I used to work there as the head of the user group. But you only scale to the billions of records. And it's kind of wild, but now data sets are hundreds of billions. We have companies that have trillions of records, you know, Grammarly, five billion new grammar things people are entering every single day. So that's hundreds of billions a year. The level of data just doesn't scale. So if your level of data goes up tenfold, you know, it's more like the cost consumption is more linear with us and it's more exponential with other solutions.

14:41The secret sauce is that since the lake house has everything in one platform with other platforms, you have to copy data, migrate data. You also have to have like two or more governance, like different usernames, different passwords. The lineage of some data is here and some over there. It's just like inconsistent. but you know really you avoid by avoiding copying of data everywhere or i mean that that's a huge deal when you get large data sets yeah that definitely makes a big difference something i'd be curious about is like how does the lake house platform compare in terms of like price and performance with you know like traditional cloud data warehouses or something yeah you know great question.

15:25There's, you know, more neutral like Gartner, Forrester, MIT report that have done their own benchmarks. Customer, you know, customers do their own benchmarks as well. And it, you know, all depends on the size of the files and the size of the data. But we, we're we are winning accounts from pretty much everywhere else. Yeah, because of those three factors. So it's not a marketing trick it's like in real life people are like ditching you know what you might call legacy systems and moving to what we call the modern data stack for all those three reasons yeah yeah i definitely see that um talking to people in the space and then something else that i have heard but i'd love to get your you know pick your get your thoughts on it but i'm wondering if you can elaborate a little bit about how uh lakehouse platform is accelerating the machine learning like life cycle.

16:21Yeah. Yeah. So machine learning life cycle, something I've done, I've been a hands-on data scientist, you know, beyond being someone like an evangelist that talks about it, like with the Chicago Cubs created all of those analytics from scratch. So the challenges I faced, you know, having been a practitioner are greatly accelerated through Databricks. So there is, you know, all different steps. There's ingesting the data, um you know called elt or etl um depending on what you're doing you're you know loading data you're transforming data so it's like this whole process uh flow and databricks has like really easy to use uh capabilities that you could give rules like when do you reject uh something super helpful when you're doing like real-time streaming like you're ingesting social media you're ingesting um you know claims or you're ingesting you know sales or something like that as things happen in real time what are the business rules to to process it and transform it but then once you have that machine learning um you know the the there's traditional machine learning on structured data only but the more types of data you can put into a machine learning model oftentimes that gives you better insights it's more uh fine-tuned to the reality to the complexities innuendos of real life so databricks does a great job taking it's called multimodal some structured data like sales some unstructured data like a transcript of a complaint call social media images and make a better forecast in there.

18:08So that's making the forecast. And then once you have the forecast, you like productionalize it. It's kind of one of the final steps. The model, is it perfect? Maybe not, but is it good enough to productionalize? Maybe put it into production, which used to be a really hard task. It would take you days to do. Now it's like click of a button. And then once you do that and you're scaled to have hundreds of models or thousands of models, it's this whole new sub-industry called MLOps. How do you have a dashboard of hundreds or thousands of models and see where the data is drifting? Know when am I going to refresh the data?

18:49Where has a model failed? Like where has the data stopped ingesting? So now you have like dashboards and alerts for hundreds or thousands of models to just understand who's using which models, what assets are being used and who. Yeah, that's incredible. That's very cool. Yeah, I think that's a big part of the solution to what was otherwise a very large problem to grapple with for a lot of people. So I think that's definitely one reason why Databricks is excelling. In a lot of ways, I definitely hear a lot of positive things about it. But something I would love to ask you, just kind of from your own perspective and in your own background and everything you've seen in AI machine learning, I'm wondering, what's a piece of advice you feel like you could give to aspiring machine learning, data scientists, AI experts, people in this field?

19:39What's a piece of advice you feel like you could give? Yeah, great question. And I do teach some college courses on data science. So love giving advice. um you know all depends on where your journey is but you know you know don't i guess my big advice is you know be a self learner um things are changing so quickly you know there's so many different websites and youtube videos your podcast to listen to so like always be learning and then like be humble enough to ask questions when you don't know you know you get stuck a lot of people are worried about like their own like it looks bad to to ask questions it's okay to fail um as long as you're learning and it's okay to not know something um you know like at my company i tell people how does this work how does that not work how do i do something um so it's a great culture but you know even even myself i've um you know had a i'm not starting out i've had a career But yeah, the advice of people starting out and anywhere along the way is just that.

20:44Keep learning, keep asking questions, and don't worry about being vulnerable and failing. I love it. That is definitely impactful advice. I mean, that's good advice that applies to anyone in so many different industries and so many different areas that people are looking to go into or to learn about. But yeah, especially I think in AI, that's really, really critical today. Something I would love to ask you about as well is, you know, what are some misconceptions about AI that you encounter in your role as an evangelist and a thought leader in this space? You know, some things that people are just getting wrong.

21:16Yeah. Well, I would say like, we're in the phase now of gen AI, which is not even a year old. So there's like a lot of like a lot of misconceptions and a lot of people are focusing on what things can and can't do. So I think like the main thing is people who just see Gen AI as like an enhanced chat bot and think that that's the only like use case that you could do or try to back fit some publicly generated LLM like ChatGPT and like use it in their company. sometimes that works but uh you know you a lot of companies are starting to realize that um you don't need taylor swift in your database you need something based on your own data for you know for certain use cases so yeah i think that's a misconception that you could use publicly available general like general gen ai models and it's like a cure-all for you know for what your business needs.

22:22And most cases, it's not. Yeah, I think that's a really, really important, definitely a misconception that a lot of people have in the space. But yeah, that's, that's some great advice. I think there for sure. Something I'd love to ask you a little bit about, of course, you are a bestselling author, you've written a number of really incredibly well received books. I'm wondering if you could give us a little brief overview or summary or pitch for, you know, your books and some that you feel like are really relevant specifically today for people to read? Yeah, well, I've gone from specific to general, but they're all like technology.

23:01So like how to do Oracle how to was one. Baseball hacks is like every baseball team has that on their shelf. It's like how to write R and Python code to like scrape data off the web and visualize it. that that was a big one one that's out of date now but that was at the time my bestseller it was the first book on windows 2000 okay and it was wild it became such a bestseller since microsoft always comes out with their own books but their book got delayed by like a month or two oh perfect yeah barnes and noble like it was the only one on the shelf if you needed to learn it um so yet it made like top 20 like beating out some for at least that window like oprah book club and everything so that was wild wild to see i went to russia and i saw it like translated without my permission uh in russian on the street it was pretty pretty fun and i saw i went to japan saw it there on the street oh that's funny when you make something that people need uh you're just gonna do whatever it takes i guess that's hilarious that's awesome i'm wondering one other of thing that I looked into and I saw about you is I know you have some patented mobile technology.

24:17I'm wondering if you can talk a little bit about that and maybe some of its implications for the future of AI and analytics. Yeah, you know, great research there. I, you know, I was talking about that like every four years reinvent yourself. So one of the ones I didn't mention was mobile technology. And I worked at US Robotics that made what was called the Palm Pilot. And people may not remember that, but it was like personal digital assistant that started the whole like mobile craze. Motorola came out with something, then Blackberry, they just made a movie on how they epically failed to capture it.

24:54And then a couple of other, Microsoft tried to get into it. You know, Apple, Samsung and others are there. But at the time it was the first business software that worked on a mobile device. So think like you can have a notebook on your computer writing code. You could do, if there are hardcore people on this podcast, like an SSH, like a Telnet into a server and write command lines from your mobile device. So how do you manage an Oracle or a Teradata database? People used to be on call and they would get paged and have to come back to the office into their home, which could take an hour, it could take 10 minutes.

25:35But here you could be out and about and respond immediately, type in a command, restart, or what have you. So that was wild. It was a company, Expand Beyond. We raised the largest Series A in Illinois in 2001, got a bunch of patents there, and got later acquired. But it was really exciting to be there at the very forefront of the mobile revolution for business. And now all of us, you know, Databricks, but, you know, everyone here, we call, like, we just had a conference and our theme was Generation AI. So we all, whether you're Databricks or not, we're all part of this, the first generation of shaping what is Gen AI and how do we, how do we explore the possibilities, make it the best we can, while at the same time, you know, limiting it in the right way, governing it in the right way.

26:30So that digital mobile database and now Gen.AI is the latest. Very cool. Yeah, it's incredible to see some of the technological advancements and shifts that we're seeing today. That's cool that you've been kind of at the forefront of each of these. And now you're kind of seeing this new kind of wave of AI. Ari, it's been incredible to have you on the podcast today. As we're wrapping up, I'd love to ask you one last question, which is based on your unique perspective, I believe you have right now in this space. What are some of the biggest changes and advancements you see coming down the pipe in like, let's say the next three years in the AI space that you're particularly interested and excited about?

27:14Yeah, well, yeah, it's been fantastic coming on here. And, you know, three years used to seem like a small time horizon, but now it seems like I have to put on my futurist hat. I used to say five to 10 years. Now I'm like, okay, three to five. Yeah, just since last November, it's incredible. There's going to be hundreds of companies coming out that don't exist now that are going to come out with incredible technology. So the main conceptual area is Gen AI. I'm just starting to see some really incredible use cases, like beyond your traditional ask a question and text, and it comes back with resources with functions.

27:57Everything dealing with real-time video, things that are going to make interactions with humans much better, humans interacting with each other much better. The other thing I see rapidly changing is the world of software development itself. So I have kids that are in college, going to go to college, and I'm like, what are you going to learn that will still be applicable probably the concepts but a lot of the first level like easy boring parts of software development are they're all going to be automated so everyone's going to if they're able to be elevated to do more and more complex so that is exciting but i just see this onrush of uh data that's out there so yeah one other thing uh i do want to pull back with the databricks is we have this credible marketplace, which is sharing of data, either open or closed.

28:56So the reason I pulled that back is to push forward that there's going to be huge marketplaces, since it's very costly to make your own trillion record data set. So I see companies where they can and should sharing data. So there's not like 50 companies scraping the same data that will be shared. But then every company will augment that with their own proprietary, like customer data. So I think that's going to be exciting. This entire like global marketplace of data, videos, of social media. And as that grows, the use cases are going to keep growing. And I want to be blown away. I want to see use cases that we're not even thinking about now start happening in three years.

29:45Yep, 100%. I think at the rate of progress we're making right now, that's not very far away. Listen, Ari, it's been incredible to have you on the podcast. Really appreciate all of your insights you've shared and your unique perspective and background. If people want to get in contact with you or learn more about Databricks and what you guys are building over there, what's the best way for them to do that? Well, Databricks.com is great. We have all sorts of like hands-on labs and demo, a demo hub, a demo center with like videos of everything. I love it if people want to follow me or connect with me on LinkedIn, just search for Ari Kaplan.

30:25I'm not the lawyer who's popular there. I'm the guy from Databricks, but I post my insights. I try to make it fun. And is that way you can kind of learn and see where I travel around the world. And hopefully I give all different nuggets of wisdom as I see and talk to people. So I love to connect on LinkedIn. Very cool. I'll end up for the listener. I'll leave a link to Databricks in the show notes. You can go over there and check out some of the really cool things they're doing. Ari, again, thank you so much for coming on the show. To the listener, thank you so much for tuning in to the AI Chat Podcast.

31:00Make sure to rate us wherever you get your podcasts and have a fantastic rest of your day.