Episode 74:

Kostas Respawns at Starburst, is Interviewed by Eric, and Reminisces About Winamp

February 9, 2022

In this very special episode of The Data Stack Show, Eric interviews Kostas about his career journey over the past decade of working in the data pipelines space. Be prepared to walk down the memory lane of data tools and realize just how far the data space has come.

Notes:

Highlights from this week’s conversation include:

 

  • Big News: podcast hits, Kostas’ career change (2:19)
  • Kostas’ career start in data pipelines (4:09)
  • The Winamp and Napster era (11:46)
  • Starting an API gateway (16:56)
  • Observing new technology from afar (23:43)
  • Starting Blendo (32:38)
  • Problems faced in architecting the product (37:12)
  • Kostas’ role at Starburst (40:25)

The Data Stack Show is a weekly podcast powered by RudderStack, the CDP for developers. Each week we’ll talk to data engineers, analysts, and data scientists about their experience around building and maintaining data infrastructure, delivering data and data products, and driving better outcomes across their businesses with data.

RudderStack helps businesses make the most out of their customer data while ensuring data privacy and security. To learn more about RudderStack visit rudderstack.com.

Transcription:

Automated transcription – may contain errors

Eric Dodds 00:06
Welcome to The Data Stack Show. Each week we explore the world of data by talking to the people shaping its future. You’ll learn about new data technology and trends and how data teams and processes are run at top companies. The Data Stack Show is brought to you by RudderStack, the CDP for developers. You can learn more at RudderStack.com. Welcome to The Data Stack Show. We are actually recording this together, in the same room, in San Francisco, which is super exciting. Kostas, I am interviewing you today, and we will hear about why that is, so I guess I’m the only one with a burning question today. My main question is actually what were the steps that led you into building an ETL company kind of at the time when a lot of interesting technology, both in ETL, and warehouses was emerging? That’s what I’m going to ask you.

Kostas Pardalis 01:02
So now I have to come up with what I’m going to ask myself?

Eric Dodds 01:09
What would you ask yourself if you were interviewing yourself?

Kostas Pardalis 01:12
I need to call my therapist, I don’t know.

Eric Dodds 01:17
This will be a very therapeutic session, I can guarantee you that.

Kostas Pardalis 01:20
Yeah, it’s a different episode, so I don’t think it makes much sense for me to talk much on the intro, but I’m really looking forward to do it and I think it’s going to be a lot of fun, so let’s just do it.

Eric Dodds 01:33
Alright, here we go.

Kostas Pardalis 01:34
Yep.

Eric Dodds 01:35
Welcome to The Data Stack Show.

Kostas Pardalis 01:38
Welcome.

Eric Dodds 01:39
This is a really special episode. So I’m in San Francisco, California with Kostas. We’re doing this together, which is super exciting, and you may notice something different about this episode: we don’t have a third person on as a guest because I am going to be interviewing Kostas today. We’ve talked about this for a while. It’s really exciting and it’s going to be awkward because we don’t even have to do intros because everyone already knows us.

Kostas Pardalis 02:03
Yeah, that’s amazing, actually. Yeah, every time you have to think like, what I’m going to do— Like, ah, no, wait: we don’t do the intros. Our guest is doing the intro, but we are doing intros before that.

Eric Dodds 02:15
Oh, that’s right. We do that. Yes, we do that before. Okay, so two big pieces of news. One, the podcast hit 20,000 downloads, which is really crazy, so that’s really fun. But part of the reason I’m interviewing you today is because you have some big career news, so share with the guests this big career news.

Kostas Pardalis 02:35
Yeah, so the news is that I’m not working with RudderStack anymore. I moved to a different role with a different company. I have started working as part of the product team at Starburst. We’ll chat about it. It wasn’t the easiest decision, but it’s quite exciting and very interesting. It’s been about a week now since I started, still very early, but it’s very interesting. The other news is that I will remain with the podcast, so we will keep doing that.

Eric Dodds 03:11
Yes. So it’s sad for me because I used to work very closely with Kostas, but we get to keep doing the podcast, which is really exciting.

Kostas Pardalis 03:18
The podcast has been an amazing experience. It was my first one. Have you done this before?

Eric Dodds 03:25
No.

Kostas Pardalis 03:26
Yeah, so it was a huge surprise for me in terms like, whatever we’ve done through this podcast and the people that we’ve met, I would never expect this to happen, so it’s awesome that I can continue doing that. That’s great.

Eric Dodds 03:44
It’s great. Okay, so first of all, let’s have a snack.

Kostas Pardalis 03:48
Ah, yeah.

Eric Dodds 03:51
We have to have a little treat here to celebrate. This is what I’m interested in talking about. This is really fun because I get to ask you questions that we’ve kind of talked about before because we’ve worked together, but now I get to do it in long format with this being the only agenda, which is great. You’ve spent almost a decade working on data pipelines, which is a really long time. I think like a lot of other of our guests, like over that past decade you’ve seen an unbelievable amount of change in the data pipelines face. So here’s my first question for you: How did you decide to work on data pipelines specifically? Was that an intentional decision? Or did you uncover a problem? What’s the story? How did you start?

Kostas Pardalis 04:45
Well, the short answer is that I did it because we had to survive.

Eric Dodds 04:52
This is gonna be good.

Kostas Pardalis 04:55
Actually, what is funny is that both me and my co-founder have blended 14 weeks. We’re working in something close to what they call congestion or ingestion platform or pipelines, even before that. The first platform we built for that was in 2008, 2009, which had to do more about what we were calling back then “metadata management.”

Eric Dodds 05:22
What were you building? That’s before the beginning of the era of warehouses, so what was the architecture that you were building?

Kostas Pardalis 05:37
Actually, it was mainly a way to monads and move around very, very structured data. That’s why I said “meta data,” not data. So the idea was that, okay, you have some assets, like images, right? You’re like videos, or I don’t know, like, digitized version of books, or whatever. And you have the digital form of these assets. But then you have like a number of metadata that describe what these assets are for, right. And you want to do like some interesting things with this data, you want to catalog them, you want to search them, you want to browse, or like use them as a way for, like a application to go and like consume them and deliver this experience to the user at the end. Now, these kinds of data models that are useful without data can become like, extremely complex, right, like, I guess everyone is aware of like, mp3, right, I can pick three. And then we have the MPEG four that was MPEG 21, the button not that far back in the past that was like a video format for that was we use it like in DVD, and then you, we give you good things like in big seven and big 21. And all these, like the actual weather, we’re trying to build like infrastructure to make information more accessible to people. And when we are thinking about information, it’s not just like, What’s the title of the video, or whatever we are talking about, like, even very low technical metadata that are needed for processing or whatever. So these schemas, as they were called the, like, data models, they were like, pretty complex. Now the question was like, Okay, how do you communicate this information between different systems, so you needed to build some kind of like, there are properties between different systems. And that’s where you had to go there and do transformations, you had to transform the data. So you have like to define it or formation between a source data model and destination data model, right. And then after you have that, you have to actually extract the data, apply these transformations and deliver this data to another system. Usually, they were like more about, like indexing the information and accessing this information systems like something like leucine, and stuff like that, or databases, like you would access it, like, sure. database. So one of the first like, the first thing that we’ve built together with photius was like a system that you could actually create these mappings between some very hierarchical or like complex and general data models between them and do like in a graphical way, like with a UI. It’s meant that you could do that even if you didn’t have like, technical knowledge, but you were a domain expert, like, you could understand the semantics. It’s always about the semantics. Even if the user wasn’t an engineer say, they understand the data model and the metadata because they are an expert in a practical usage of that data, or whatever the downstream thing is. Interesting, okay. Keep in mind, back then you didn’t really have a discipline of data engineering. You have in the best days— I don’t big data engineers that were trying to figure out how to do MapReduce, and you also have the advantage. So if you wanted to somehow make systems that were operating on the data level, you had like to create these kind of topics and transformations, and you need the chairman to who has like the domain knowledge do that. So that was like the first thing that we read.

Eric Dodds 09:16
Did you build, like, was that a company?

Kostas Pardalis 09:18
That was popular Funkadelic research that we were doing.

Eric Dodds 09:23
Really? At a university? Interesting.

Kostas Pardalis 09:27
It was so successful that actually there was an opportunity to commercialize that technology. Some institutions and commands that they came and they were asking, how to buy this thing and use it like internally, and that’s pretty much like the time where I started thinking of, maybe I would like to start the business.

Eric Dodds 09:46
Interesting. How did the companies who were interested in buying that technology find out about it? Did you publish a paper or something?

Kostas Pardalis 09:57
There were a couple of papers but the main way that these technology goats are known in the European Union, it was because it was funded by the European Union.

Eric Dodds 10:08
Interesting, right? Yeah. So, yeah, the research is funded by the European Union.

Kostas Pardalis 10:13
So usually the way that you does that is they try to create, like consortiums around the projects that involve many different countries and different organizations there, because it’s one of the ways that one of the tools of they have to build, let’s say, a common European market or like, culture or whatever. So that sense, they were like, Okay, we were also spending money and going, like in conversation, like talking about these things and like trying, but nothing that anything close to whatever you call, like a go to market strategy, or whatever.

Eric Dodds 10:46
I think it’s pretty hard. I’m not an expert, but I think it’s kind of hard to get IP out of the academic environment and into a startup.

Kostas Pardalis 10:54
I would assume it depends on the country. Greece is not the place to do that, for sure. But it’s not easy. Yeah. Like the legal framework around that stuff. I mean, it because he used also include, like, actually, the IP belongs to the European Union, right? Because they funded the research. Yeah. Which means that it belongs to the people of the of Europe, which means, like, you have to make like Openshaw anyway, let’s say. So what you can do is like, you can sell services around that, like, that’s the Yeah. Now there are, I think, today, at least some ways that you can do like, create, like a spin off or something like that. But by Justice, like a very long time since I was involved in the academia and business models around.

Eric Dodds 11:39
Do you miss it?

Kostas Pardalis 11:40
No.

Eric Dodds 11:43
Okay, I have other questions, but when you were talking about the different types of MPEG files, I couldn’t help but think of Winamp. Did you run Winamp?

Kostas Pardalis 11:55
Oh, yeah.

Eric Dodds 11:56
Because you could you could really do a lot with the audio quality and dialing that in and then the different skins for Winamp. I had such a modded Winamp. It was awesome.

Kostas Pardalis 12:09
Yeah, we know them. Then you had like, it was like, Napster? What was the—

Eric Dodds 12:13
Napster, but that was like the— Winamp was, or at least like— Man, I’m trying to remember. Winamp was like the audio player. It was very highly customizable.

Kostas Pardalis 12:25
Yeah, and I think it was like downloads.

Eric Dodds 12:29
Yes, but it changed over time. But then it became the place to just get download illegal files, right? Yeah. And it was like a ton of music. And then, you know, video came like, I mean, it was crazy.

Kostas Pardalis 12:41
Yeah, actually, the first experience that I had with like, downloading music and stuff like that was through IRC. I think it was by the end of the 90s beginning of zeros, there was a lot of— and that’s one of the main ways that I got exposure to the American culture, because I’ve met many American people. There were these channels that you would join. And people were building boats that they were like fighting servers, and they had like commands. So they would expose the whole, like mp3 list through these boats. And you could download them through this free, amazing experience. But what is great is like the level of like, automation with this technology that we’re talking about, like 20 years ago, right? And we had a lot of, I remember, it’s funny. We had like many problems like with spammers or to join the like, oh, yeah, spam the tunnels. I remember that I created at some point a book to help control the spammers. So I, they’re,

Eric Dodds 13:48
Would it detect them?

Kostas Pardalis 13:50
Yeah, so I don’t remember exactly the day of. I’ll have to think a little bit about it, but there was like a pattern with the IPS and also the reverse DNS and like, what kind of reply you would get. So, okay, sometimes you will also kick some people that you shouldn’t kick out. But most of the spammers, they felt like this pattern with their DNS and the script or the both what was doing was like running these reverse DNS getting the response saying like, and based on that and some heuristics around it, like was kicking you out.

Eric Dodds 14:28
You know, it’s interesting like thinking back to that era. This is flooding memories with Winamp and Napster. The amount of innovation that people achieved just trying to like pirate huge amounts of music. Back then, gigabytes and gigabytes of music seemed huge.

Kostas Pardalis 14:51
Oh, yeah, and like it was taking a lot of time. Most of the digitizes things and three like it wasn’t like ripping is Be like, oh, yeah, it was like he didn’t read the vacation.

Eric Dodds 15:05
Totally. Yeah, I remember the I went to like a large state university here in the United States. And the, you know, students ran the IT department basically. Right. And I can’t remember who someone made friends with one of them. But they set up a private, like file share system on the internet. And it was unbelievable. I mean, it was like everyone’s media, like, I mean, like, people taking hard drives to them to upload the file system like, amazing.

Kostas Pardalis 15:41
Yeah, I was doing the same thing. Now the university. The other thing that was very fun, it was back then what we were calling like, LAN parties. So because, okay, we’re a big fan of like, playing Quake Arena. So, one of the servers that I was hosting at the lab that I was working back then at the early zeros was, like, around, like, an Arena server. And we would. I mean, that was not allowed, actually, but in weekends or, like, Friday nights, or something like that, because we have the keys of the lab, we would get people to bring their PCs in there and have like a LAN party and play with people from the US, for example, because only the university had, like, enough latency to play competitively with someone from the United States while we were in Greece, right? Yeah. So

Eric Dodds 16:33
yeah, it was, gosh, it was nice. It is amazing. The creativity that can come from people motivated to circumvent rule.

Kostas Pardalis 16:42
Yeah. Instead of starving.

Eric Dodds 16:48
Okay, thank you for indulging my little detour there to discuss modeling out linium. Okay, so you’re at the university, you started thinking about starting a company? Was Blendo the next thing?

Kostas Pardalis 17:04
Tom pointing both me and like, later microphones or like 40s, were like, Okay, we’re done with what we’re doing here. We for that age back then we had saved like, some small amount of money that we could survive, like for a couple of months or two years without working. So we’re like, Okay, let’s stop doing whatever we’re doing and tried to figure out like, if we can, okay, obviously, we had no idea what that means, like, all of that. But during that point, we were working, we’re making money, but we were working with like freelancers. And sure, just doing like engineering work, right? So and we’re talking about like, okay, all these things are happening, like in the country that it’s like, the opposite of healthy environment for the printers that we don’t have is as part of our culture, like, please, right? Like, it’s not something that you are encouraged to do. So we were trying to figure out what to do. Actually, we didn’t want to do exactly what we were doing other university like research as part of our research. So we started at the beginning, by building a technology around API management. So it was like an API gateway with a graphical interface, again, where you could do like some create action, like an API gateway and route all your API calls through bots.

Eric Dodds 18:20
I don’t know why, but it surprises me that the first sort of two big—I know you probably built tons of stuff—sort of product type things that you built had a big UI component.

Kostas Pardalis 18:36
Oh, yeah. Because back then technology was all about how we can build products that can be used by non technical users, right? And even if you were like a developer, you would still, the idea was like, Okay, let’s try and move away from the terminal or like, using, I don’t know, like something like, the experience that you get with emacs or VI or whatever. And that’s like, the time that I like, all the IDs here, you know, what developed? So I think it’s, it’s both like a cultural thing, and also has to do with the maturity of the industry, right? Yeah. Like, we’re talking about the periods where, okay, being like a software engineer, it was special, but it’s not like everyone was trying to become like, a software bug thing. Okay. So it’s not like we caught that many people to do these things. And the main way that the industry was trying to solve that was by creating tools that are very easy to use. So the UI component was quite important. And we were still in the process of like, trying to figure out actually like, how we interact with machines anyway. So

Eric Dodds 19:45
yeah, that was I think, I’m trying to remember the exact era when like a bunch of the FTP, like desktop apps came out, you know, to just make that yeah, like a drop off like you know, whatever a drag and drop experience. Yeah, then just make FTP, like a lot simpler. Yeah.

Kostas Pardalis 20:03
100% Like it’s, it makes sense. I think that’s how, okay, we, you can solve everything with, like, in an efficient way at least. But now that I reflect back to then I think it was like very natural to try. Yeah, thinks this way. So yeah,

Eric Dodds 20:18
you leave all the front well, you left on the front end debugging to me.

Kostas Pardalis 20:23
I was I was always a very, very bad engineer, developer, whatever when it comes to from the inside, like, that’s,

Eric Dodds 20:33
well, that’s why it surprised me. Because like, you don’t gravitate towards that naturally. No. Yeah.

Kostas Pardalis 20:38
I hate it. I kind of do it. Like I’m almost blind. Like, I think it’s very funny when I work with designers. How many details like I’m missing over time. It just counts. So you put stuff

Eric Dodds 20:50
interesting. Yeah, you did a pretty good job.

Kostas Pardalis 20:54
Okay, we just got like pretty good designers, I guess.

Eric Dodds 20:59
Fair, fair. Okay. So API gateway.

Kostas Pardalis 21:03
So we have that we have like a demo, we’ll do the demo around this. And then we’re trying to figure out how to I don’t know what to do next. Like, okay, we have technology. How do you turn this into a business? That’s a completely different kind of problem. Now we are talking about like, 2013. Keep in mind that we are post-2008 which means that the rest of the world is recovering but Greece is a mess financially.

Eric Dodds 21:38
But the data space now has some of the tools that become what we know. The name brands of today are just starting to be built.

Kostas Pardalis 21:53
Correct. You have snowflake starting. I think they started in 2012. You have…

Eric Dodds 22:03
Redshift was earlier than that, maybe?

Kostas Pardalis 22:06
Yeah, I think Redshift was out there, so the whole cloud data warehouse revolution started. I didn’t remember probably it was a little bit later. Maybe anyway, around the time data breach was there. I mean, spark was a thing from like, 2009, I think and at that point, I think they also started the company. Okay. It was still like a very hot do oriented era, right? Like, it was a cloud there, Ira, whatever. And most of the business noise was around bi so you had Looker, Tableau? Tableau, I think was like older.

Eric Dodds 22:42
I were there. around for a long time. And yeah, Crystal Reports. You remember that? I think it’s still around actually. IBM’s high. Yeah, like, well,

Kostas Pardalis 22:51
that’s awful, I guess. Yeah. Still rounds. What else? subnuclear we had baby scoop data moved analytics charter yo, like all these components, like started, like pretty much at that point. Yep. And it’s like, also the period where you know, Y Combinator started becoming, you know, like, the brand name that we know. Yeah. So I remember like, I think, No, they’ve like the CEO of he was one. I think it was like one of the first like, cohorts of YC hearts back then. So it was like early even for like, Y Combinator. Yeah.

Eric Dodds 23:28
Okay, so a question for you: What was it like to look at the emergence of both these technologies? And then also, I guess you could say, I mean, the startup ecosystem in Silicon Valley is nothing new, it predated Y Combinator, of course, but you were looking in on the emergence of things that are now wildly influential, both tooling and startup, but you were looking at that from the outside in Greece that’s still recovering from the economic crisis. What was that like? Being here in the States, even though I went through that on the East Coast, it still felt— My guess would be that it maybe felt closer. I don’t know. What was that like?

Kostas Pardalis 24:21
Actually, it’s a very interesting question about your asking because back then I remember that I was telling to myself to how late I was for the mobile revolution. Right? Hmm. So this whole thing with like, Mumbai, hops, and like all the hypergrowth that you saw communication, like oh, yeah, I missed that. Like I was just late enough. That didn’t make that much sense. Especially for someone who was like from from my blue. I was not here, right. Yeah. Now, I don’t think I could imagine back then. That’s like what The stuff that we were talking about and starting to build back then after like 10 years would be the equivalent of what mobile was back then. Right. So yeah, I thank you for the question because it just made me realize like, the is that life is a cycle at the end, like, you know, like, you just need to be at the persistent enough and you will get like to you will get another opportunity. I don’t know, like back then I couldn’t understand anything to be honest. Like I was in an environment that was like, super, like, depressing from a financial something like Greece was like, in 2013 wasn’t like that bad. But we were reaching the peak that happened in 2015, where we had the capital controls, right? Yeah. Yep. So you were living in a country that everyone was depressed, like, both literally and metaphorical in the market in the financial market, when it comes to venture capital, that’s like almost non existent in Greece. And the VCs that existed were just very nice, kind of like, investors that had many to do with ecommerce, more b2c kind of sure business model. And whatever I could be alive was through the internet, and like reading Hacker News, and the grounds and like, stuff like that. So sure. It’s not like I have any kind of sense of what Silicon Valley’s at the end outside of, you know, like, whatever the popular, the media likes, tried to communicate to everyone outside. So yeah, I had an idea, actually. But at some point, we had to move forward. Like we because okay, we had like the technology, we had, like the stupid thing that we build. But we had to figure out a way to turn into the business. We didn’t know how to do it. We knew that like we cannot find people to help us in Greece, because nobody knew how to do that. So we started looking for opportunities to get some investment from abroad, and see what happens, right, I think that have applied to Y Combinator like four times, probably, I don’t know. I was applying to everything that would give you money. That’s why I was saying I was just trying to survive. I never got to the point where I had an interview or something with Y Combinator. Then we were trying to find money.

Eric Dodds 27:31
This is still for the API gateway?

Kostas Pardalis 27:33
Yes, yes. We were trying to find people in Greece to invest. No luck. Also, we didn’t have any kind of network, like, because that’s important. I mean, no matter what we say, investment is a social process. It takes time, and people need to trust you. And you need also to trust them in order like to engage in this kind of relationship. So okay, you don’t just go open a door getting signed, like, Okay, give me money, and you get your shirt. Especially if you’re someone who you haven’t done anything in the past, like, you don’t have any kind of like, track record.

Eric Dodds 28:07
Learning how to raise venture capital isn’t a class that they teach in academic research. I mean, they probably do now.

Kostas Pardalis 28:18
When you have been educated as an engineer in the classical way of engineering, you think of everything in a very Boolean way, like everything is like black or white. It’s either I have a solution, or I’m still working to find the solution. Right? And that’s like, what, like a big shift that you have to make in order to become like an entrepreneur, or like a product person, or anything that has to do with like, interacting with people out there. And sharing defines, like solutions for the problem. And that was like one of the biggest problems that I had the beginning because like, I couldn’t realize that so yeah, we had built something but I wasn’t feeling like like I was feeling super embarrassed like to go and present this thing because it felt like broken so it was hard. So we kept looking for money. The third co founder joined the company had some experience he was working for a while for Motorola like Motorola in May I think their first headquarters were like in these are anyway, they had like some very strong ties with Israel, which makes sense also, because of like the engineering background of Israel and Motorola being like in telecommunication and stuff like that. So he knew a few people like in Israel back then Israel started using a lot the whole brand name of startup nation and we have stand ups blah, blah, blah, and all these things and trying like to expose like the country in a positive way to the rest of the world. And he was like, Okay, why we don’t also try to go to an accelerator or like something like that to Israel. And it was something that to be honest, like, although he’s there and it’s very close to greet like it’s hour and a half Like, same time, it’s not like it’s very close. I never thought about it because Greece, like, during that time didn’t have like very strong ties like with Middle East and Israel specifically. Oh, like, okay, sure, why not? I mean, sounds like anyone else’s accepting for another like an amazing experience because what happened was that Microsoft had an accelerator in Israel, Microsoft, Israel had an accelerator there. I send an email there, it’s getting come back to them. They reply, like, okay, what are you doing? Like I’m saying, like, dada, we’re building like a new API management platform broke or whatever. They’re like, Oh, it doesn’t fit with our theme for this year. But I’ll connect you with some other people that was like, Wow, amazing. I mean, that’s nice. Anyway, so by sending me from one to the other, which is a very, like, beautiful thing that like the Israeli cultural hub, we managed to find this mall and micro VC there with some young and very crazy people that were like, Okay, let’s give like some money to these silly Greeks and ask them to come and live here for a couple of months and see what happens. And that’s how I want us to raise our first capital that was like, so such a small amount of monocot. Like, if I felt the number, I think people will be like, okay, like, what the fuck is he talking about, and we took that money, and we went to Israel. And by the way, the period that we were discussing and the terms, Israel was like, fighting, like, it was crazy. That was like another crazy experience that really exposed me to the real world. And like, so that like, you know, like thinks, good things, bad things can happen, like at the same time, so the fight stops on Sunday or something like that. On Monday, we were flying to Tel Aviv. And we went there and stayed there for like, a couple of months. I think it was like a little bit more than seven where we tried to build the company. And that’s when we moved from the ABI thing that we’re doing into what Glenda became with the pipelines. And it was mainly the result of feedback and our inability to sell the API platform for testing many different reasons. Back then I was thinking that it’s just a failed product. But it wasn’t like a felt prod it was like the combination of like, timing and people was wrong. Like, there were problems like, we just couldn’t sell it. And it was like a very bad time in the market. Also, there was a consolidation happening without staff, so there wasn’t that much interest.

Eric Dodds 32:32
So you started Blendo. When did you pivot from the API gateway to Blendo?

Kostas Pardalis 32:38
In 2014 we incorporated the company so we would get it was August, September went to Israel. And we tried to sell this thing for like six months, we kept saying no, and at some point, we were like, Okay, we need to do something different. If we want to move forward, we need to pivot. And so we had a very interesting conversation with the Product Manager from IFTTT back then who came to IFTTT through an acquisition, like he had a company that got acquired by them. He was in the States but the Israeli ecosystem always had like very good ties with cm Valley. So one of the investors that we had knew the guy and send him an email, and he was like, Okay, this is what these guys are doing. Like, can you give them some feedback? And he was like, Okay, looks fine. But I’m more interesting about the data behind the API’s. Oh, interesting. And not the API’s themselves. Because the API’s themselves, like, I’ll manage them once and be fine. Like, I’ll create like a digression there. And that’s it. Like, it’s not something that they have to manage that data about often. But the data behind these API’s is something that I’m missing as a product manager right now. Like, I need all these information from all these different tools to go and do my analytics. It’s like a pain in the ass right now to do like, I got access to data. So we’re like, that sounds interesting. close to what we were doing before, in a way so we had some experience on that on the technical perspective. So like, okay, let’s be bold and let’s do something that has to do with data. Again, it’s a bit different than what blender became at the end. But whatever started that day is what like actually I would recognize and identify as Blendo.

Eric Dodds 34:32
Interesting. When you say it was different, because Blendo Blendo really was, in the early days, head-to-head with like Stitch Data, Five Tran, basically a direct competitor.

Kostas Pardalis 34:45
That’s what we became. The problem with us was that we didn’t really have access to the market. Okay, like we got this information like from this person that said, like we couldn’t even have like direct access to him because from what I understand The inventor was a little bit embarrassed about us. So he wouldn’t connect us directly. Yeah, seriously. Like, he was like, no, no, no, I cannot let us go and talk to this person directly. So we had no idea. We just have, let’s say direction, right? And then when you have a direction, a science fiction. I mean, you build like finding things. And we did that. And because we were afraid to go to markets, we were trying to avoid it, right? So we’re not selling it, where it’s a point where we put it, like, sustain ourselves anymore. And we were like, Okay, that’s what we have right now. We put the website out there. And we asked for people to subscribe and pay and someone paid. Like, what the fuck like, wow, my $20 Being an entrepreneur is easy. And then yeah, like, I think the rest is like a bit of history in the sense that then you set let’s say, all the things in motion that will push the component the product to some kind of direction, right? So after that, it was much easier, but Right, like, I think at the end, it’s like super important to go to market as soon as possible. Because especially if you are a first time founder, okay, if you are more experienced, probably like you can constrain yourself from things like stupid thing. And like stupid things, like being two founders is the motto on what you should be building, but going to market, I think it’s like probably the most important things that you can do. If you’re like, early, if you’re like a first time founder.

Eric Dodds 36:32
Fascinating. I’m so glad that I got to hear the story about the founding of Blendo and what you build before that.

Kostas Pardalis 36:39
Yeah. And the crazy thing is that, like, all these things happened, like the through the period with capital controls. Yeah. And because we were selling outside Greece, because of course, like we didn’t have a market for ETL. Like, doesn’t have even to date on and we were forced to incorporate the company in the United States, because we had to take money from Israel. We had the bank account in the United States, and we didn’t care about the capital controls.

Eric Dodds 37:04
Interesting. Okay, well, I know, this is a big surprise, but I think we’ve gone longer than we intended. Two more questions for you. We’ve spent a lot of time just hearing the story, which is great, but what’s the hardest technical problem that you faced in architecting the product? Does something stick out where, whatever. It was just very difficult to build or difficult to scale or something kept breaking as you were building Blendo?

Kostas Pardalis 37:39
That’s an interesting question. Because I don’t think like I’m gonna give you a technical answer to it. I don’t think that we had technical problems, we have human problems mainly. And I would say that, like, this is probably the case, like with most of the things that I have seen so far, like, okay, we’re not sending James Webb to, to the springs, right? Like, okay, we’re solving like, hard problems, the main problem starts, the main problem starts with, when you have to scale, I think most of the problems happen, because of how hard fast you have to move, and what kind of rate those you have to make there. And but again, the problem is not the technology, it’s the interaction with, like people and how you can, let’s go back to what I said about engineers like you, you keep thinking like in black and whites, right. And you have to think like that, like, you shouldn’t change that, if you want to be like a good engineer. Now think that on one side, you have like, you’re an engineer who’s building a product that knows that it’s not the right thing to scale. And then you have a company that has to scale in order to survive. So, of course, like, your go to market things are going to push towards that. Now, this balance between the two and making sure that, like, things are not going to collapse, it’s like the hardest part of building something new. In my opinion. I failed many times in the like, it’s I think, I still keep failing. But I think the problem is on the technology itself, like, the technology that we have, actually, we’re very spoiled. Like, we have so much capacity to do like you we live in a period where if like, we like we can scale just by burning money and throwing hardware into the problem not even optimizing the source code necessarily, right. Yeah, the code is not the problem. I mean, unless we are talking about like, I don’t know, solving, like cancer or something. It’s it’s the humans that complicate things and make things interesting. That’s right.

Eric Dodds 39:44
Yeah. So that aligns so well with a theme we’ve heard on a podcast throughout the episodes, I don’t even know what count we’re on but which actually feels very authentic because, technically, you know, you’re I mean, you’re a host The show, but you’re a guest on the show today, and you’re from a different data company. So it feels very authentic that you’re repeating the theme there. It’s not about the technology. It’s about the people. That’s true. That’s true.

Kostas Pardalis 40:11
But yeah, it isn’t. I mean, you, you build business for people, right?

Eric Dodds 40:17
Yeah. Okay. So that that was the that was one question. I said I had to What are you doing at Starburst?

Kostas Pardalis 40:25
I’m doing products. So it’s not that different than the things that I was doing at router stock for a long time. I’m focusing more on like my focus, because Okay, the producting. There is like bigger than the team at rather stack. My focus will be more on the core technology source Starburst is built on top of three. No.

Eric Dodds 40:43
We had the Well, Josh anniversary. Yeah, the CEO on the show. So

Kostas Pardalis 40:48
the company is actually bringing to market like a commercial version of the open source version of three No, and why focus is more like on the core technology, three know itself, the query engine, the optimization of the query engine, some stuff like materialized views, and some product use cases on top of that. So that’s my focus, at least this week.

Eric Dodds 41:14
That’s pretty good for your first. Yeah, awesome. Well, I’m, I am personally very excited to just hear from what you learn there. And I think it’ll bring a great new perspective to the show. So I’m super excited for you. And I can’t believe I didn’t say this. Congratulations. It’s an exciting new adventure.

Kostas Pardalis 41:31
Thank you. Thank you, thank you, it was, as we say, like, I’ve been like building pipelines infrastructure for quite a while I needed some change, and actually, like, torture my brain with, like, new problems. And so I wanted to go a little bit like, farther down the data stack, like user to store ads and databases them. So a query engine, like Threema was like might be an opportunity for me. And query engines, like, I feel, especially now with the emergence of data, lakes and Lake House architecture are going to become like even more important, and are like very hard. Also, engineering problems. Yeah, if you want like to optimize them. So there’s a lot of opportunity for me to get in front of like new problems and see how we can productize this problem. So it’s, I think it’s going to be like an interesting journey.

Eric Dodds 42:23
Well, as we say, so often, we’ll have to have you back on the show and learn more about your experience working on a federal All right, well, thank you for joining the data section with Casa tonight in the same room in San Francisco. I’m so excited for you as a colleague, and more importantly, a friend and I’m most excited that we get to keep doing the show together.

Kostas Pardalis 42:48
100% Yeah, I am really looking forward to see also what’s going to happen this year with the show like it’s been like, it’s been like a crazy one here for the show. I mean, we announced today that we did like 20,000 downloads, right? Like the show’s job there for like, almost a year. Yeah, it’s crazy.

Eric Dodds 43:07
It is why I’m Yeah. Okay, this is I’m just gonna ask you another question. You know, because like the debrief is weird with a host interview. What is it like to have, like a clean inbox, just starting a new job. That’s, I mean, that’s happened, I’m sure to like our listeners into like me a couple times throughout my career. And it’s like crazy. You start a new job, and you’re like, your calendar is so open, and your inbox is so empty.

Kostas Pardalis 43:34
Yeah, mom’s like, too stressful. Polish food, like something is wrong. The first day, I was like, no, like, something’s wrong.

Eric Dodds 43:47
He spent 10 years like an overloaded inbox. And

Kostas Pardalis 43:51
yeah, I mean, one of the things that I did was like, I went through my personal email, but I never clean it. I clicked it. So just to feel that I have many emails. Yeah, yeah. The good thing is that doesn’t last for too long. I mean, the good and bad thing, but it’s a very weird sensation. Let’s that’s true.

Eric Dodds 44:08
Yeah. All right. Well, that was. That’s our outro. We hope you enjoyed the show, and we hope you enjoyed gasses not being in the same room. We’ll definitely do this again as we have opportunity, and we’ll catch you on the next show. We hope you enjoyed this episode of The Data Stack Show. Be sure to subscribe on your favorite podcast app to get notified about new episodes every week. We’d also love your feedback. You can email me, Eric Dodds, at eric@datastackshow.com. That’s E-R-I-C at datastackshow.com. The show is brought to you by RudderStack, the CDP for developers. Learn how to build a CDP on your data warehouse at RudderStack.com.