Douwe Maan: If you're trying to build open-source software, you are really limiting yourself if you only see it as a distribution model or, oh, we'll get some free users and then we'll convert some of them later on. We really see it as core to the way the company is built, where your users are not just users of a product, but actually semi team members who feel like they have the same access to information and same access and ability to give feedback that your actual full-time team does. Eric Anderson: This is Contributor: a podcast telling the stories behind the best open-source projects and the communities that make them, I'm Eric Anderson. I'm joined today by Douwe Maan, who is the CEO and one of the founders of Meltano, which I think is the name for both the company and the open-source project. Is that right, Douwe? Douwe Maan: Yep. That's right. Thanks for having me, Eric. Eric Anderson: Yeah, totally. This is going to be a great discussion. There's a lot we can cover. So I want to jump into all that, but first you should tell our listeners what Meltano is. Douwe Maan: Yeah, probably a good place to start although the history is very interesting too, and we'll get to that in a bit. So Meltano is the open-source data ops operating system. What that means is that we are trying to bring software development best practices such as DevOps, version control, code review, and CI/CD into who the data worlds, where data teams are setting up their modern data stacks by combining seven-ish different tools, mainly drawing integration lines between them, and then ending up with a data stack that does everything from pulling data from sources to visualizing it with dashboards or allowing data science work to be done. So where Meltano comes in is that SD data stack has kind of exploded in terms of amount of available tools, something has gone missing, which is this connective tissue that allows teams to still treat their data stack as a single software development project, essentially, and allows the entire thing to be taken through software development life cycle principles, such as end to end testing, isolated environments on a team member's local machine or in staging and in production. Douwe Maan: And it's Meltano that provides this foundation underpinning the data stack, where you can bring in these different components which are open-source or SaaS and running in the browser, install them on top of your Meltano project, so that you end up with your entire data stack declaratively defined within a repository where a team can collaborate on it as if it were any other software development project composed of a combination of off the shelf and custom components. And Meltano aims to bring the qualities of DevOps into the data life cycle, by providing this layer that allows it to be treated as one. And it provides integration between components that is difficult to set up by an individual data engineer on an ad hoc basis, thereby making it easier than ever to get started with the modern data stack. And then also to evolve as the modern data stack changes and you want to try out new tools because you're not locked into your choices of the past and the manual lines you've drawn between various components. Douwe Maan: Meltano improves, increases choice by allowing people to build their ideal data stack with Meltano providing some continuity so that five years from now data teams can still come to the same place to reason about their data flows through the various systems while the specific components used might look nothing like what the project was originally set up with today. Eric Anderson: So you're a GitLab for data ops, is that too- Douwe Maan: Ah, yes. Eric Anderson: Not, no? Yeah? Douwe Maan: Sort of. So we can talk a little bit about history because that's why GitLab is relevant. I, myself, I started working at GitLab in 2015 when GitLab was just going through Y Combinator and raising its seed funding rounds. And I was there as an engineer originally, and then as an engineering lead for a number of years and in 2018, the data team at GitLab and in general, people kind of looking at other opportunities for open-source software in the broader industry, realized that there was a need for data tooling for data teams built around a lot of these software developments principles, and based on open-source components, which we believe for many reasons that we can go into later are going to take over the world of data just as they have software development. Eric Anderson: So GitLab is building this DevOps pipeline and they're specifically we're kind of building every component of it in open-source, right? Douwe Maan: Yeah. So good points. I can a little bit more color there. So GitLab started out and this still today, this open-source end to end platform for the entire software development life cycle. So that means entire software product teams, whether you're talking about the product manager or the UX designer or the QA engineer or the people managing the deployment on top of Kubernetes, can come together in one on platform that does everything from issue tracking to a virtual control and code review to automatic security testing, deployment, everything you can imagine in one place. And at some point the realization was had that a similar opportunity existed in the data space. Although there's a number of differences, one significant one being that GitLab is one tool that does it all. And of course has great integrations with external tools, but is mostly still a one size fits all solution where you can integrate alternatives that are more focused on a particular part of what GitLab does if you have specific needs in one of these seven or so steps of the software development life cycle. Douwe Maan: In Meltano's case, because we were looking at the data space where five to 10 years from now, there were many of these end to end platforms that a company would adopt wholesale and would do everything for their data needs, this space has evolved really rapidly and on every layer of the stack or every step of the data life cycle, there are at least five, 10 competing tools. Some more established and also new ones being founded and funded every single day, meaning that a single platform that tries to do it all is no longer realistic. So when Meltano started out originally, the hope was that we would be able to leverage open-source components that already existed to some point, while also building our own first party solutions for areas where we thought that open-source solutions were lacking. But since Meltano's founding in 2018, now almost four years ago, things have changed a lot. And Meltano has ended up focusing exclusively on being that operating system layer that allows all of these components to work together seamlessly instead of trying to actually build every component ourselves. Eric Anderson: Got it. So interoperability is kind of foundational more so here at Meltano? Douwe Maan: Absolutely. And leveraging existing open-source technology, as opposed to trying to build our own answers to areas that we think are underserved. Eric Anderson: Fantastic. Okay. So you already gave us a little teaser at the story, which is what I'm excited to hear. So you were in the early days of GitLab and presumably you were excited about software development lifecycle at the time. And at some point turned on to this idea of the data platform. Douwe Maan: So it wasn't me who turned onto the idea of the data platform. Within GitLab the decision was made that we're going to bet on a couple of opportunities that we think might be a similarly large open-source business opportunity as GitLab had been. And one of these because of those clear needs of the data team who were looking for great data tooling and finding themselves in a world that looked like the software development tooling world of 10 years ago, where everything you're doing is done in production. There's no concept of making a change locally and verifying that it works before putting it out there. And they thought they're deserve to be better tooling for data teams and also for software engineers that are kind of bridging that gap and starting to do more data work. And on the other hand for data teams that have seen with envy, the kind of tools available to their software development counterparts and want some of that. Douwe Maan: So in 2018, Meltano was founded inside GitLab and I joined the project in the autumn or the fall of 2019. Well, where to go from here? There are so many places we can dive into the story. What do you think Eric? Eric Anderson: Yeah, there are a lot of places. I mean, at some point we should cover when you've realized that this could be a separate entity. So within GitLab, the idea was, we're going to have this incubation effort. And what's now called Meltano was one of those. Were there other things? Douwe Maan: There were some envisional ideas being played around with, but it turned out that a lot of those did end up making sense as part of the larger GitLab suite of tools. So some of those became features at some point. Data was one where we realized that the market and also the needs of the market were different enough that we would at least start out building this as a separate tool, not as a feature on top of GitLab with a potential path towards inclusion in GitLab. Although that's something that we ended up deciding against when the company was originally spun out. So I'll just cover a little bit more of the background of the project all the way through, from 2018 until now. So the goal was always a single platform that can do everything for the entire data life cycle, built around software development best practices and open-source from day one. We realized really quickly that a lot of different companies were building internal data tooling and not really sharing it. Douwe Maan: So there was tons of duplicate effort taking place. And with GitLab, we've known from day one that collaborating on tooling is how you get the best tools for your users by not just getting their feedback through support or product management routes, but actually being there with them in the issue tracker, in the trenches essentially, where they can see every line of code you write, every single feature you prioritize and throw in their opinions. So from day one, Meltano was much the same, but we came to realize over the course of 2019, when we realized this end-to-end vision where Meltano could really do everything from data extraction, pulling data out of various systems and showing you a dashboard presenting some of those data points, that we realized that the data space was unbundling really quickly. And there was not a place anymore for one tool that does it all. Douwe Maan: It was also very difficult for people to start contributing because you would kind of need to adopt Meltano wholesale instead of saying, "Well, I'm pretty happy with half of my data stack, but I want to try Meltano for this and that." So by the end of 2019 when I joined, we were working on this V1 of Meltano that would prove the end to end story. But unfortunately we also came to realize in early 2020 that we weren't really seeing the traction with the product or the project in general, that was necessary for GitLab to continue investing it at the level that it had been investing in it with a six person full-time team. At the time, myself being engineering lead with four engineers and a general manager. So in early 2020, the decision was made that in order to give Meltano the best chance for eventual success, essentially the head count was getting reduced from six down to one, thereby extending the runway sixfold. Douwe Maan: And I was chosen to be left on the project by myself, essentially with a mandate to try to turn it around over the course of 2020 or GitLab would be forced to reallocate resources and call this a learning exercise, but not something that ultimately delivered something of significance, open-source or business value. So in early 2020, I was in this position where Meltano had been around for a year and a half. We had a bunch of code written. We had some people in slack that had been following the story, but hadn't really become active contributors. Douwe Maan: And from talking to a lot of the people who had shown really early interest in Meltano, it became clear that one of the most compelling things we had built was actually a layer around an open-source technology called Singer, which is a standard for open-source data connectors that can function as extract and load pipelines so that you can build a little executable, which can be essentially be written in any language but it's usually Python, to pull data from a source and then a separate little script that loads data, or writes data into some kind of destination whether that be a data warehouse or some other web based, cloud based tool or a file format even. And you can combine these two connectors to essentially create an EL pipeline. But in the ecosystem of Singer, there were literally hundreds of these different connectors, but no open-source tooling that would really allow you to orchestrate these pipelines and run them in production, manager configuration and allow teams to develop their pipelines with some of this software development life cycle advantages pulled in. Douwe Maan: So in Meltano, by trying to leverage existing open-source technologies to the extent we could instead of reinventing the wheel, we had already adopted the Singer standard for connectors as what would power Meltano's EL extract and load functionality. So basically from one day to the next, based on a lot of this early feedback that I got, I decided to change the positioning and marketing on the Meltano website, away from what we used to call ourselves then from data to dashboard in 60 seconds or whatever, but then open-source ELT platform to really specifically focus on those integrations with Singer, with DBT pretty much the leading data transformation tool out right now. A really, really great tool, great community, super happy to support it in Meltano and Airflow to handle the actual orchestration of the pipelines and the schedule at which they should be run and the error messaging. Douwe Maan: And Meltano was essentially that glue, that intermediate tissue that made you actually want to use these three tools together instead of coming across them, but not really knowing how to tie them together and make it something production ready. So over the course of 2020, Meltano started picking up usage really quickly by people from the Singer community who were looking for better tooling to run their Singer powered pipelines. Eric Anderson: Your experience there is not what most technical leads on a project are doing. User interviews, marketing tagline, optimization, but it seems like what the project needed at the time, was this kind of a new muscle for you or? Douwe Maan: Yeah, it's interesting. Before that, of course I had been a software engineer for a long time. I knew that some of these more people activities came a little easier to meet and to many other engineers. So it wasn't necessarily a surprise that there was some belief that I could pull off a lot of what this project needed to happen in 2020 to really be turned around. But every day I ended up putting on a different hat. One day I was product manager, the other day I was community support, the other day I was recording a talk or a podcast. And then there were days where I was just writing documentation and articles all day long. And days where I was doing interviews with the handful of users who had found us. Douwe Maan: Although, of course we were in a very lucky place where we had a slack community already with hundreds of in it at the time, just through that early GitLab associated interest where a lot of this getting user feedback wasn't even so much an explicit thing I had to do, but something that just was already naturally part of the way we worked, where the issue tracker is in the open, slack is in the open. I would write monthly blog posts kind of updating the community on the progress and there were always people sort of come in and sharing their opinion. And when it became clear that that was explicitly what we were looking for in order to find where to go with this project, people were super happy to be part of that process. Douwe Maan: And that's really one of the lessons also that GitLab learned very early and that we have also taken to heart now that Meltano has been spun out, which is that if you're trying to build open-source software, you are really limiting your yourself if you only see it as a distribution model or, oh, we'll get some free users and then we'll convert some of them later on. We really see it as core to the way the company is built, where you're trying to involve your users in all the processes and all the decisions that come to the company and how it builds its product and how it prioritizes. Where your users are not just users of a product, but actually semi team members who feel like they have the same access to information and same access and ability to give feedback that your actual full-time team does. Douwe Maan: And that goes as far as having our handbook be completely in the open, live streaming our weekly kickoff meeting on Monday. Of course, we have office hours every Wednesday where we kind of talk about the top of mind features and include the community in those conversations. And it always leads to better features and better first iterations of things we want to do than if we had welded it too long internally before finally showing it off to the community. And that's also a big lesson of how, that first year or so of Meltano is where we were really kind of chasing this conviction we had about the need for an end-to-end open-source data platform didn't match the reality that we found ourselves in when we really started talking to users every day. And that's a mistake we won't make again. Eric Anderson: There's so much to unpack there, certainly the dream of chasing a platform only to realize that to be relevant, you have to build a thing. You have to start somewhere and there's generally not the end to end everything thing I think is very common and it sounds like you got there quickly, which is big. And now that having solved that thing, it's given you the opportunity to now be a platform again. You can kind of expand from there to solve more problems. Douwe Maan: Exactly. And it's a different platform than what we wanted to be originally because originally we thought of a platform as in one massive thing that kind of does it all and might use some open-source technology, but that really particularly became an operating system, an actual foundational layer on top of which everything else sits. And coincidentally sort of the infrastructure, the architectural design of Meltano's code base, when we realized, okay, we need to support Singer. We need to support DBT. We need to support Airflow, became what is now the operating system, because Meltano was always about combining very different tools that might be written in different languages and might have completely different ways of being configured or deployed and building those abstraction layers to allow them to work together seamlessly. Douwe Maan: And it turns out that that is actually the most compelling thing we built more so than the idea of an intimate platform to do it all, because in the modern data stack, people are essentially building their own platforms to do it all just by picking six, seven different components. But it's just Meltano that makes it finally feel like one product again. Eric Anderson: Yeah. And then the other thing I heard from that earlier story of yours was your acknowledgement that there's open-source, which you Google says they do with Chrome and Android, but really they just kind of publish the code when they're done working on it. But you're talking about the whole new level of open that GitLab encompasses. I don't know, open planning or open everything, what you would call it, but you're better at the taglines clearly than I am, but you're right. That forces people to do a level of customer user orientation than they would normally otherwise do. Douwe Maan: Yeah. And one thing I think that made GitLab so good at this and something in which it was different than a lot of other companies is because GitLab was open-source project, purely an open-source project for literally a year or two, three before the company even came around. GitLab was started in, if I'm not mistaken 2011 in the Ukraine by the Dmitriy Zaporozhets who until recently was GitLab CTO, because he or his clients in Ukraine had a need for an alternative to GitHub that could be self hosted. And that would allow its users a lot more control over what the tool did. And of course, developers are special in that they are capable of building their own tools. So you're kind of depriving developers of something if they cannot tweak their own tools. And you're more likely to get really, really awesome tooling if developers can put all their love and their effort into it. Douwe Maan: So for years, GitLab was really quickly catching up with GitHub in terms of feature set and everything up until the point where Sid Sijbrandij, now CEO of GitLab realized that there was a business opportunity around this project to fund it's ongoing development, and also build some kind of enterprise business model around it, either through a hosted version or enterprise functionality, which is exactly what Meltano is doing in the near future as well. But the reason why this is different is because from day one, GitLab was built by a team of hundreds of people around the world. And when 1, 2, 3 of them started getting paid, fundamentally nothing changes about the way that GitLab was run as an open-source project. Everything still happened in the same issue tracker, whether you were getting paid through the work or not was kind of an arbitrary distinction. There were people with merch commit rights who were part of the core team, externally to the company and those inside. Douwe Maan: And because this team was so widely distributed around the world already asynchronous communication was a given. Remote work was a given. None of these were necessarily decisions that GitLab made. They came from the fact that it was an open-source project. And since from day one, this community of hundreds of users always felt as much as part of the GitLab team as the tiny little team of 10 people on the company's payroll, including myself at the time was. From day one, Meltano and GitLab have always been built with their users instead of, like I said earlier, open-source being a distribution model, more than anything. Eric Anderson: Got it. Totally. And in the same way that GitLab had its early days outside of the commercial period, Meltano had early days attracted some users and you kind of had to take it from there to its own entity. You're to the point now where you've built the ELT open-source Airflow Singer Meltano setup, but you've moved beyond that. Right? So what happens next? Douwe Maan: Yeah. Great question. So in early 2020, when I decided to focus purely on ELT for the time being that meant focus specifically on three plugins supported Singer, depth and targets being connectors, DBT for transformation and Airflow for orchestration. And having proved the concept of this foundational layer, this operating system pulling together these technologies, we have now earned ourselves the right essentially to start adding in support for more and more different types of data tools and also more and more specific data tools. So in the very near future, hopefully still to be released in February, we will be adding support for great expectations. For example, one of these data validation tools that is becoming a part of many modern data stacks, we are adding support for Superset, which is an open-source dashboarding and visualization solution. We are looking into adding support also for alternatives to Airflow like Dexter and Prefect, so that we can really show our users that Meltano is not ELT. Douwe Maan: Meltano is not singer and DBT and Airflow, Meltano is this foundation layer that allows all of these different tools to be tied together and allows data teams to design their own ideal data stack without Meltano being too prescriptive with regards to the specific tools being used. So we articulated and we started writing about this vision for Meltano as the data ops operating system in October or so. So we're only about three months in. So in the past three months, a lot of our time has still been going on the side of the ELT functionality and making sure that we have a really, really compelling offering there. And now in coming weeks, we'll start releasing more and more of these abstractions that allow other data stack components to be run on top of Meltano. What we've already started shipping in the recent weeks also is data op specific functionality that brings some of these software development best practices and principles to data. Douwe Maan: So you think about isolated environments, configuration environments so that a local user can run the entire stack and be connected to one particular data warehouse while another user can be playing with the staging instance to find out any last minute bucks. And then there's always the production copy running that the actual CFO is using to create his dashboard or what have you. These isolated configuration environments is a fundamental feature in Meltano where you need some centralized configuration management layer because all of these different components of the stack have their own systems. Some of them have text based config. Some of them have config toward the database and there's really no way to take your entire data stack and say, "Now we're in staging. Now we're local. Now we're in production." Douwe Maan: Unless you have something managing the configuration of all of these components, being one of these Meltano features we've released. Similarly end to end testing, allowing you to detect whether a change in your extractor configuration would accidentally break a transformation or whether a change in a transformation would accidentally break a dashboard that depends on some kind of table or column that is now missing. This is the kind of testing that can be built into Meltano because it has full control over every aspect of the data stack. And these are also recent features that we are kind of using to introduce people to some of these DevOps and software development best practices. Eric Anderson: I feel like I'm apparently four months behind the messaging, because I kind of had put you as more of an ELT solution and clearly there's a lot more going on. And maybe that becomes really apparent in the February with the new launches. Douwe Maan: Our new website literally launched in the last week of January with that whole new data ops OS messaging. So we've been starting to talk about it a lot internally to our existing community, but it's not a surprise that if you've just come across us on the internet so far, we'll have come across mostly as an ELT solution. But the important thing to realize is that Meltano is not ELT. Singer and DBT and Airflow are arguably ELT in orchestration, but it's Meltano that brings them together and makes them better than the sum of their parts. And Meltano will be known for that last aspect and bringing in those data ops qualities more so than the specific functionality, because ultimately a Meltano powered stack is what you make of it because we want to support all opponents and give you the choice to find the ones that work best for you and prevention from being locked into a best technology decision that doesn't align with your needs anymore. Douwe Maan: So the hope is for Meltano to be able to grow with companies as they go from tiny startups to massive enterprises, as well as to grow with the data space as tools get replaced, and people want to do stuff site by site, and they don't want to feel this massive barrier to trying out the new stuff because their current stack is so brittle and held together with duct tape. That just seems like a very daunting effort. Eric Anderson: I do feel like the data open-source world has always been a bit brittle and maybe in part, because there's some ad hoc nature to it. Things can kind of be triggered once and sometimes they work and sometimes there's less dependency. It's not as real time production user facing. And so we accept amount of brittleness, but it's good to see that it's, I think as enterprises lean on data more and more we're going to need real robust systems. Douwe Maan: Absolutely. And what we also see is that, of course, every organization in the world could benefit from making more of their data. Realizing the full potential of all of that data that's out there and the insight that are essentially waiting there to be found, but the data stack as it stands today, because it is comprised of however many components that takes some manual effort to first select and then also configure and integrate and deploy, the barrier to entry to starting with the modern data stack is higher than it was in five years ago where you could just buy one massive intimate offering and kind of be done with it. So a really big part of our challenge is also bringing modern data practices to organizations of any type, including a lot of those that might not have been in the market for some of these big proprietary non open-source solutions. Douwe Maan: I live in Mexico city and I see here every day, extremely talented teams and programmers and startups doing really, really groundbreaking stuff either for Latin America or just for the wider world, but just by the nature that these countries work, some of these US-centric SaaS tools are just out of their budget and will be for a very long time. So better open-source tooling, more accessible open-source tooling, and a data stack that can be set up by a single engineer instead of requiring a team of expert data engineers. We think it's going to make a massive difference for the world's learning from data in general. Eric Anderson: One thing that would help me understand the new operating system positioning some is, in the past when we've talked about these open-source data systems, there's been the distinction of being in the data path or not. Like Airflow kind of orchestrates other pipelines where spark and other things are kind of carry the data, even though they both have these dags inherent to them. Do people create dags in Meltano? Are you in the data path or you kind of more of an orchestrator or do those names not really apply? Douwe Maan: Great question. So I wouldn't say we're an orchestrator like Airflow is. Meltano doesn't orchestrate workflows and it tries to use existing tools instead of reinventing the wheel. But it is essentially an orchestrator of components almost similarly to how Kubernetes and Terraform are ways of declaratively defining various aspects of your software application stack. And then the tool translating that into the needed compute resources and making sure all the tools are consistent in terms of configuration. So Meltano is an orchestrator more in that sense, but because we want to make it really easy for people to build pipelines that leverage multiple plugins that have been brought into Meltano. So that could be Singer for EL, DBT for transformation, great expectation for testing. We do have in Meltano, a native kind of pipeline concept, which will likely turn into a deck concept down the line, but these are currently still executed by Airflow with Airflow calling out two Meltano. Douwe Maan: Meltano runs CLI commands which then can take these different components, inject their configuration on the fly and still rely on the orchestrator when it comes to error handling, scheduling and all of that stuff. But by defining some of your pipelines using these Meltano fundamentals and then essentially using the orchestra as a swappable engine also makes it easier for people to migrate from one tool to the other, or to use tools side by side, if they want to tap into some Dexter or Prefect functionality, for example. Eric Anderson: I should look at the docs before I say something like this, but it sounds a little bit like a Terraform for your data pipeline, where I can kind of describe in a normalized fashion here's a pipeline, it's going to rely on Airflow for execution. And this pipeline has these abstractions that we're going to inherit from Singer, table, descriptions and then I can tell it to run. And later on, I can swap out to a Prefect or Dexter or if I wanted. Douwe Maan: Exactly. Terraform for data stacks or a package manager for data tools are both apps comparisons that cover some part of what Meltano offers. Absolutely. One big difference though is that a really core advantage of Meltano is that if you have reached a point where you know, that you want to start using some of these software development best practices in your data workflow or in your data organization, you kind of need something that brings them all together and allows these tools to be treated as one unit. And that's essentially what Meltano does in making data ops a reality finally, but it does that by treating these different components, not as purchasing decisions or as products that certain people in your team spend half their day in, but rather as components of the team's own ideal data stack, which is more akin to a software application than to anything else. Eric Anderson: The stuff coming in February feels very reminiscent of a software application and the need for Meltano that I could define this thing that I described a moment ago, and then I could pass a configuration for one execution, production configuration, and I could pass it a staging configuration for another execution. And I could have those in kind of a, I want to say CI/CD, but I don't know what words I would use for data, but yeah, some kind of continuous deploy setup where every change, I might ask it to do an execution on the test configuration. Douwe Maan: Yeah. And your CI/CD comparison also your hesitation for whether that's the right word is interesting because when you're working with data, of course, the data itself kind of flows through pipeline. So you want to validate the shape of the data itself, and you want to make sure that there's no failing queries somewhere in your transformation pipeline, but at the same time, these transformation definitions and these extractor definitions and these dashboards definitions are themselves also version controllable assets. So the first thing we're trying to do is make building dashboards and changing transformations and configuring integrations with other systems, part of that life cycle that should be version controls. And then kind of orthogonal to that, you also have the validation and the testing pipelines for the data as it flows through the pipeline itself. But where the advantages that just DevOps alone, version controls, CI/CD, isolated environments alone bring to data is also in reducing individual team members, kind of fear of changing things in tools they don't intimately understand. Douwe Maan: Because if you're always working in a production environment, you're not going to dare touch a button, unless you're super, super confident what it does. In the software development world, you can just let a junior or an intern do whatever they want to your code base, because they're just doing it on a local machine anyway. And if their local machine application breaks, so be it. Then it would never make it through CI/CD and code review and into production, unless it all works. And increasing people's kind of appetite for experimentation and ability to have multiple people on the team make tweaks in stuff that would traditionally fall on the other side of the fence is part of increasing the velocity and effectiveness of data teams. And we're really kind of on a mission to merge those worlds of software development and data engineering, by letting both of these people realize that they have far more in common than they have of dividing them. Eric Anderson: Good. Let's go back. There was one topic I wanted to cover around the transition of bringing Meltano out of GitLab. We went over the reason why it happened and the result of it, but I'm curious how that went over with people within GitLab, how that went over with users. How do you roll out such an announcement? When did you know you would be kind of taking on the new role and how did you feel about that? Douwe Maan: So, as I said earlier in early 2020, I took over the Meltano project. And during 2020 things really started taking off to the point where even though up to that point GitLab allocating more budget had been not a given, kind of contingent on more traction and success. In early 2021, we were at the point where GitLab had all confidence, allowing me some more head. So immediately I hired Taylor Murphy from GitLab's own data team to be our head of product and data. And AJ Steers from the Meltano and senior communities to be our head of engineering, at the time still inside GitLab. And literally within a month after they joined in conversation between Sid and I, Sid GitLab CEO, we came to the conclusion that GitLab as an organization of like 1500 people, all working on one and the same product and marketing strategy and sales organization, except for these three people living off in this little land called Meltano, the big, big organization just was not set up to be as lean and as efficient when it came to serving the needs of tiny little startup like Meltano. Douwe Maan: And within the organization, Meltano was always seen as this cool little thing that might be something one day, but most people in the company also didn't feel very connected to. And even in the community of Meltano, people knew that it came out of GitLab and of course that association was very much a positive one. GitLab has a stellar track record when it comes to treating its open-source community and still being able to build a successful business around that. But people really loved seeing when we make that decision to see if we could raise seed funding and spin out. Us being given the opportunity to really spread our wings and start growing this thing with VC backed exponential growth, instead of within GitLab having to sort of compete for resources and compete for attention with the rest of the organization. So in January 2021, I was still by myself. End March 2021, we were three people. And then by May, we had raised funding. And then by June we were spun out and the team had grown to eight people. Eric Anderson: What a year. Douwe Maan: Yeah. And now it's January 2022, and it's just insane that a year ago I was still doing this by myself and now we're a company. We have a broad, ambitious vision. We have 2000 people in slack. We have 11 people on the team right now, 11 more job openings on the board, check out meltano.com/jobs if you want to be part of this journey or just join our slack community, if you want to give us feed on what we've been working on and how we can address your own data needs. Meltano.com.slack, thanks for allowing me to plug those. So, it's been quite a year for sure. But also in those first months after coming out of GitLab, we really realized how much of the machine we were kind used to. Nothing in the sense of oh, the machines long is down. The machine helping us. I didn't have to think about payroll. I didn't have to think of about so many little things and random inbound emails from payments that are not going through or in all kinds of accounts payable and invoices, et cetera. Douwe Maan: And just simple things like okay, figuring out what our compensation strategy is going to be. Figuring out all kinds of things. But it also gave us the opportunity to really create our own unique values, heavily inspired by GitLab's, but with some twists and to figure out how we wanted to do things based on in some of their team members' cases, five plus years at GitLab, having seen so much of the good, but also some things that we thought we could do better or differently. And the goal was very much to not make the same mistakes. We're going to make mistakes for sure, but they're going to be different rocks we'll hit our toes on. And it's been super exciting also to going to have a rebrand, new logo, brand new websites. We're organizing all kinds of community events, which is something that we just didn't have to bandwidth or the budgets for originally. Douwe Maan: And the confidence both on our side and GitLab's and that of the investors is that we are in a position to, in some sense, replicate the GitLab story, but also very much with our own twist in building a stellar open-source product with a thriving community and thousands and thousands of users around the world. And then at some point, figuring out an open core business model with most likely a hosted edition, proprietary enterprise tier of functionality all the while treating the open-source community extremely well and realizing that we literally cannot do it without them. And it's not even about using them for something. It's about building really, really awesome tooling in extremely close collaboration with their end users, which was always the thing that attracted me about GitLab in the early days, as well. As a software developer getting to work on software for software developers. Douwe Maan: Your job is essentially to scratch your own itch and make your own job easier. And now with Meltano we're seeing very much the same, where data engineers love this opportunity to build tools that make them more efficient and where they can bring in all of their ideas and share those with the rest of the world and ultimately level up the entire data profession. Eric Anderson: Fantastic. So tell us who's the best fit for Meltano? What kind of organization gets really excited about if they have the right tools or something? Douwe Maan: Great question. So we have seen interest across the board from tiny little companies with maybe one technical person on their entire team, suddenly tasked with setting up a data stack and realizing how daunting that can be if you haven't been doing that for years, who find that Meltano is the easiest path to a production ready data stack that still allows them to use some of those software development advantages they might have gotten used to. And at the same time enterprises or just large tech companies with well equipped data engineering teams who are empowered to make their own tooling decisions and realize that there is something missing in their data stack and that they do want some sense of continuity in a single place that brings together all these tools. But what we're seeing on the one end is people who are setting up new datasets from scratch doing this work, but also people who realize that they have a data stack that they set up at some point. It's kind of hard to change anymore. Douwe Maan: It's hard to try out new components, hard to swap out stuff that aren't working. And they realize that in order to enable themselves and their tech stack their data stack to evolve in the future, they need to first go through that transition of defining it in a way that builds on top of the operating system. So that from that point on, they can much more easily bring in new tools and try out new stuff without being afraid to break things. So it's software developers who have never done data work before and were kind of set to see the state of data tools where you're still kind of working in production in like the early days where you're FTPing into a PHP web server and making changes to files on the fly to seeing in the browser, if they're broken or not. Which is very reminiscent to the state of the data stack today in many places. And data teams that are envious of the awesome tools and capabilities available to their software development teams. Douwe Maan: If you fall anywhere on that range, which is a very wide range, there will be something to like in Meltano for you. And we encourage you to come check out our slack community, join 2000 more people who are making this vision happen with us, or to see if there's a job at Meltano where you can do this work full time. Check out Meltano.com/slack or meltano.com/jobs for either of those. We also have a event coming up from the 17th to the 24th of February called Love Tep Fest. Tap here referring to Singer Teps and targets these connectors for extract and load pipelines that can run really easily on Meltano. So if you'd like to learn from our community of 2000 and our team and the wide, wide singer community, how to build connectors for new sources that can run on top of Meltano or just to build other types of integrations for the Meltano platform, come join us. There is some swag giveaways too based on how many new connectors you can get done. Douwe Maan: So whether it's a new connector you're building or an existing connector you're trying to migrate over to our Meltano SDK, you're more than welcome from the 17th to the 24th of February. More details to be found on meltano.com. Eric Anderson: Thank you so much Douwe for coming on the show today. Super exciting place you're at, at the moment. Douwe Maan: Yeah. Thanks Eric. Thanks for having me. Really fun to talk with you about this and talk to someone who also knows the inner sense to open-source angle so well. Eric Anderson: You can find today's show notes and past episodes at contributor.fyi. Until next time, I'm Eric Anderson and this has been Contributor.