Data Journey with Chris Tynan (DistroKid) - Data & AI in large-scale music distribution to Spotify, Apple, Google and more

Speaker A: Foreign. Welcome to the next episode of the Radio Data Podcast. Today our guest is Chris Kynan, who lives in London and works as a data director at uh, Distrokid. Uh, Chris, uh, welcome to our show.

Speaker B: Hello Adam. Good to be here as always.

Speaker A: Uh, let's start with the introduction. Uh, so Chris, could you tell us more about you and about your company and your role at Distrokid?

Speaker B: Yeah, of course. Uh, great to be here. So, as you mentioned, I'm from London or I live in London, I'm British. Um, yeah, I'm a director of data at Distrokid. I'm a long term data and analytics professional. Um, Azimu and I work together at Spotify, what feels like many years ago at this point. Um, but throughout my career I've done a whole bunch of different data and analytics roles. So not just in the music industry. Um, district kid where I work now is a music company. M. I'll talk about it in a minute. Uh, but I've worked in a whole bunch of different things. Uh, my very first job out of University was joining UBS, the UH, investment bank in 2008 during the credit crisis, which was very interesting. Um, so I've had many years in the finance industry looking at algorithms, back when we called them algorithms and before we called them machine learning or AI. I've worked at music at Spotify, now at Distrokids, and I also had a stint working within the British government, um, both within the Home Office. Um, so our kind of like state Department that covers immigration and visas and national security and other things. Uh, and quite a, um, a few years at the British competition regulator that cover competition law and consumer protection. So particularly at this time, we're in right now, very interesting intersection of technology and regulation. Super fascinating. Um, I've been at Distrokid now, my current employer, for about the last four months. I'm quite, quite new to the role. Uh, but it's been really, really exciting. I'm really, really happy to be here. Uh, Distrokid is a distributor of music. Um, it's a wonderful independent distributor of music. It's a tool or a service rather that any individual within the world can go to distrokid.com you can go there right now. You uh, can pay what is relatively a pretty small amount of money. Um, our um, plans start at about $2, US$2 a month and for that $2 a month you can get all your music that you produce uploaded to all the various streaming platforms in the world. So Spotify Apple, Deezer, Pandora, you name it. Like, we will put your music onto those services. Distrokid does a lot more than just that. But paring it down to the bare bones, that is the service. Uh, one of the, uh, fantastic parts about Distrokid, and one of its kind of unique selling points is that if you as a musician upload content through Distrokid to Spotify, for example, you will keep 100% of all the royalties that you make on that music. So periodically we get the royalty information back from Spotify and we distribute it to you. We don't take a cut. The only way in which we're making money is off that subscription fee. So it's an extremely transparent business model. It very much does what it says, um, and we think it's very empowering to artists and, um, helps put more great music into the world. Mhm.

Speaker A: And can you tell more about your role at Distrokid? You mentioned that you are new at the company, but still we would be very happy to hear how your daily work looks like.

Speaker B: Of course, yeah. So, um, I joined Distrokids to be the director of data. Distrokid is not a particularly huge company given the amount of music we put into the world. We put a huge amount of music, uh, I think we may even be the largest distributor of music globally, if you kind of count it by the volume of content that we put onto the various services. Um, given that we don't have that many people, um, and the company has been around for quite a while. It's founded by an incredible individual called Philip Kaplan about a decade ago. And we've grown a lot in recent years, especially around, um, Covid. You know, a lot of people were suddenly at home. They didn't have anything, or they didn't have as much to do, or some people were in that position. Uh, and as a result of that, you know, we saw a huge increase in people creating music, you know, in their bedroom or whatever it might be, and using our service as a way to get that out into the world. Um, when I joined, uh, really the role was to combine certain elements doing data and analytical things within Distrokid already, but that were not at that point unified under one team. So we had, uh, a small team of analysts. We had three analysts, two of which were very much focused around the finance and the marketing world and were reporting into the cfo. Um, we had one product analyst who was part of that team, but very much focused on like the product world and all the product analytics. That we do. And we also had two data scientists whose roles were very much supporting the operational side of what we do. So I mentioned, you know, I alluded to the very large volume of music that comes through our service. There is a huge amount of work that we put into reviewing that content, making sure that content is appropriate to go onto the various streaming services and also dealing with, to put it politely, sort of bad actors who would like to do uh, things like streaming fraud or upload copyrighted content or claim to be Beyonce when they're not actually Beyonce. Uh, so we have two data scientists there and they were very much working with the operational side of the business. So it was about bringing these groups together under one function, um, which I'm really excited about. I have an experience both in data science, machine learning and on the analytical side. And uh, around about the time that I started we also hired a data ah, engineer. Um, so we already had people within the company who were doing uh, like we had a DBA team and we've got a lot of people who are doing data engineering type things. Um, but the new person that we hired at the same time as me, uh, so far his role has really been to kind of uh, help build data pipelines and infrastructure that really can empower those other two teams, the analysts and the data scientists to do their jobs even better than they're already doing.

Speaker A: Yeah. So you already shared a number of use cases that uh, you are working at Distrokid. Could you maybe first tell us what type of data do you have and how you use that data to implement data science and machine learning algorithm?

Speaker B: Sure. Uh, it's quite a broad question. Uh, we'll see what we can do.

Speaker A: Um, so what is the most important data set that you have?

Speaker B: That really depends on the use case. So if I were to sort of split it into two parts, we've got the kind of standard um, data that you would expect from any kind of sort of consumer facing company. So we've got you know um, transactions and payments data. We've ah, got usage data of the site, you know, capturing activity, just what people are looking at on the site, um, looking at things like funnels and all the kind of standard B2C metrics that you would expect a company like us to be looking at. And then we've got the Distrokid specific data, uh, that's probably more interesting here. So we've got all the kind of uh, every time a user is uploading a piece of content we obviously capture that. So we've got the music, we've got the artwork, we've maybe got lyrics, if lyrics are provided for the songs. And then in terms of like the positive reinforcement, feedback loop, we work very closely with our streaming providers, the digital streaming providers, the DSPs, Spotify, et cetera, um, our partnerships with them. As part of that, they provide us with feedback about users or content. But really it's content, not users, because they don't know fundamentally who our specific users are. They just see the artists. So they could come back to us and say, hey, we notice that like you've uploaded this song, um, maybe this song shouldn't be on our service because the album art um, contains an inappropriate image. And you know, we take that back and we feed that back into our process and we say, okay, we will build a better model to identify that type of content and make sure it never gets onto the service. The fascinating thing about this is that the, the services only see the content that makes it through our uh, filter already. We do a lot of work before a piece of content ever gets sent to a uh, streaming service to make sure that it is appropriate. So just uh, because it's a pure volume game, a lot of stuff in an absolute sense does make it through the funnel. And like stuff does unfortunately end up on Spotify that we then later remove or Spotify removes it. But compared to the amount of stuff that we capture before it even makes it to the service, it's much lower. So we do do a really good job in that realm. And that is a combination of um, a degree of machine learning algorithms and other forms of heuristic algorithms, but also a huge operational effort and a lot of human in the loop interaction to make sure that things that we might identify or flag, you know, a huge amount of them are still reviewed by a human being before they are either decided to go onto the service or not onto the service.

Speaker A: What is the percentage of the music that eventually is uh, delivered to uh, music streaming apps such as Spotify or Apple?

Speaker B: Um, definitely the majority of our content that people upload makes it onto the, onto the service. So it's, it is the majority, um, quite a reasonable majority of it. Um, but there is still a, I'm not willing to commit to a specific number just because of the sensitivity of it, but there is still a very substantive amount of content. You know, we're not talking like, we're not talking like 99% makes it through, but we're also not talking like 50% makes it through. You know, we're somewhere in the middle of that, where there is enough content being captured every week. Um, and it is super. The role that the operational teams play in reviewing the content along with the algorithms is super critical. You know, we've learned through the history of the business and the length of time that we've been here, we've developed a team that have got extremely good domain specific skills, uh, quickly reviewing, processing content and like being able to identify quite quickly along with tools that we've built. Identify. Okay, this isn't something that should go onto the service. And as I say, they are um, flagging quite a lot of things every, every week also some stuff gets flagged and it can be relatively benign and it just requires the, the artist to change something. Maybe uh, the lyrics are not formatted correctly or maybe they just need to change the album art because they don't maybe realize that the art is not in a format that is appropriate for the streaming services. Um, we try to provide that feedback as much as possible, but um, sometimes we have to kick it back to the user and say, hey, please make a change here.

Speaker A: So I assume that there are good actors, uh, who would like to submit their music content to uh, streaming apps. But from. So, but for some reason it's not uploaded because maybe they, they use wrong lyrics, maybe they use wrong images, something that they are not aware of or something that is not intentional. But you also mentioned that you have bad actors who claim to be an artist who they are not, or maybe they are using the content which uh, which they don't really own. So, so what do you look at to identify those bad actors?

Speaker B: Yeah, the term sort of bad act is definitely true. I mean you can be within the realm of sort of music distribution and putting your music onto streaming services. There are a few different ways in which you can kind of be breaking the rules, so to speak. Um, things like uploading, uh, content that is, you know, under copyright by somebody else is a very sort of clear violation of copyright law in terms of service. There are solutions in place out there that do audio fingerprinting that, you know, we're able to leverage so both within our own catalog, but also third party solutions that allow us to take a piece of audio that we receive and run it against, um, you know, databases of, of existing tracks. We can flag things that way. Um, there are things that are within the middle ground. Um, but fundamentally if somebody is trying to be a bad actor or so to speak, what we probably mean by that is somebody who is seeking to monetize, to generate revenue from the streaming services, um, in a way that is not uh, is in some way sort of artificial. So they could either be uploading content that doesn't belong to them and trying to sort of claim it or they could be saying, you know, I am such and such a person when they're in fact not.

Speaker A: Um.

Speaker B: But probably the thing that's more prevalent and more talked about within, certainly within sort of the music um, press as well as a little bit more mainstream at the moment is this concept of streaming fraud where this is where somebody puts uh, music onto a streaming service and usually it will be their own music. Um, so it won't be copyrighted music, it will be their own music. Um, but then through various mechanisms they seek to artificially inflate the streaming volume for that track in a way to try to generate revenue. And this is not a new invention like I'm sure. Adam, you and I both worked at Spotify. We remember the very famous Wolfpack um, story. Wolfpack were uh, and are an incredibly popular famous band who released an album of silence and then had their fans repeatedly listen to the, the tracks on the album, I think they were 30 second tracks. Listen to them on repeat to generate royalties. It was kind of ah, it was both a PR stunt but also an illustration of the um, nuances of the payment model of how streaming services eventually pay their artists. But coming back to sort of distrokid in our content, we work with the streaming platforms quite extensively to root out and identify that type of content. So we've got various mechanisms internally um, to try to mitigate this happening. Um, sometimes we're caught in a sort of X post set of activities. So if we do find somebody who's been doing this or if a streaming service does tell us that this is happening, um, we have the ability to kind of look at patterns of behavior. You know, are other people using the same credit cards, like that sort of thing? Um, basically to, to work out exactly uh, what the patterns might be. We do have a series of models that we've built uh, that help us identify when we believe users are engaged in this type of content based upon a variety of historical factors. Um, we think this performs reasonably well. But also it's still a very much a, uh, there's a, there's an element of human oversight to make sure that we are indeed catching the right people.

Speaker A: So having this said, what do you think about generative AI? Because uh, because this can definitely help you to build better product. I think that you might already have some ideas how genai could be used at Distrokid, but it can also help uh, bad actors to generate some music or um, do something that, that can help them actually to cheat the system or uh, abuse the policies.

Speaker B: Yeah, yeah, it's a very interesting point. I mean I think of it like many other leaps forward in technology. Um, it enables a new set of creative expression. Uh, it also, when you enable new sets of creative expression you also like provide some ways for people to try to sort of take advantage of that or use it for sort of nefarious means. I think that's no different than other things. I mean, you know you could say the same about Photoshop. Photoshop is an amazing tool that allows people to do great things. It's incredibly useful but also it allows people to put fake images into the world. And the, the problems exist very often exist more not with the tooling in itself but more with how the tooling is used by somebody and uh, the types of actions that people do with it. And you know we seek to um, monitor and um, respond to users behavior rather than the tool itself. Um, I don't think we take a particularly strong view at all about the nature of how the content is generated. I mean I think within reason like there are certain types of content, like pure sort of complete white noise type content or content which is essentially for lack of a better word sort of garbage. Like no one, no human being would actually listen to this. And I think the streaming services are responding by kind of saying we don't want this type of content on our service. And we I think also are respecting that. And you know we, we do seek to be providing that to be sort of handled on our end. I do think that it is super interesting. We don't like have much of a handle on oh, did you use AI to kind of build this type of tool? Um, I think there's slightly more like uh, the slightly more kind of borderline interesting questions around things like deep fake um, voices. So you know these kind of sounds a lot like voices, uh, and the degree to which that intersects with copyright law. I think we are very much an active like watcher of that debate and very interested to see where it goes. But I think when it comes down to it at the end of the day probably the big issues and the things for us to catch are uh, probably there before generative AI and will be there after generative AI which is still people trying to commit fraud and they will do that with or without generative AI. I'm not convinced yet that we've seen much evidence that generative AI uh, is something that will enable or uh, is enabling sort of fraud or streaming fraud at a scale that didn't already exist beforehand.

Speaker A: Definitely you have a lot of data so you can measure um, the impact of generative AI, the positive impact and uh, the negative impact. So uh, let's focus maybe on the positive impact because you also mentioned uh, earlier that during COVID uh 19 pandemic time, uh you uh, saw that there are more and more people creating their own music content and uploading that to uh, music streaming services. Which totally makes sense because for instance people saved a lot of time on uh, commuting. They uh, designed their uh, home office studios, uh, but now they can also use Gen AI as an instrument uh to help them actually to generate even better music even if they are non professional uh, artists. So this was, this was recently mentioned in a podcast, uh, where Gustav uh, Soderstone, uh, one of um, key people at Spotify, he mentioned that Gen AI will be used by people in that way as a kind of instrument thing which lower the barrier of entry and people can uh, produce uh, ah, music in easier way. And do you also see the increase of the music content since for instance uh, November last year when ChatGPT was announced and this uh, whole AI revolution uh became democratized?

Speaker B: I think we've not seen the kind of uh, sort of shift, you know, like m, like in the way that you might be sort of envisaging or describing sort of. It's not like we've suddenly sort of hugely spiked, uh, and we're on a different trajectory than where we once were. And I think that's not a huge surprise. I think it will take time um, for these generative AI tools to find their ways into kind of very specific use cases. You know, we've come a long way obviously in terms of the um, sort of foundational models and technologies that have been built. But either those technologies need to get packaged into a product that is very relevant for a specific use case or you have to look at incumbents within the music creator tools space and say okay, you know, at some point these um, companies are going to integrate more of these types of features into their products themselves. You know, that's not going to happen overnight. I'm sure that will be something we see more of in the next two to three years. So I see a sort of, I'm expecting a more gradual increase in uh, in that than anything else. I actually think it's a very positive advancement. I think as you've described the ability, um, to get access to different sort of. Maybe it's different instruments or different tracks. There's other things as well, I think. You know, you could look at mastering. Um, Distrokid has got a wish. We, since the start of this year we've got a mastering product, um, which means if you upload a track and you have. And you sort of use our product to do this, it's. It's a sort of a separate product that is built separately at the moment. But if you use that product and pay for it, then we have a service that allows you to sort of master your track and hear what it would sound like. That sort of thing. I think more tools within the creative workbench for musicians is really interesting. I am not a musician myself, so I don't want to overstep my, my knowledge of what those tools look like. Um, but I think that sort of stuff is super interesting and I'm sure we'll see more and more of that type of thing to help musicians produce, you know, just better music or even just to kind of um, generate ideas or to generate the creative process. You know, whether that's musicians working in tandem with AI, whether that's sort of lyrics or ideas for sort of seeds of tracks, that sort of thing I think is super interesting. I think the other thing that we're seeing um, sort of contemporaneously is trying to make more and more of this type of tool and capability, uh, available directly on mobile devices. Like, if we think about where the overall kind of growth, um, globally in music is coming from, obviously a lot of it happens in Western markets, but we've got really evolving an interest in music markets in places like Nigeria where, um, it's a lot more of a mobile first country and a lot more people are just doing things on mobile devices. Certainly at Distrokid, we see a lot of content that is uploaded through mobile. We don't know if it's like, was created purely on someone's iPhone. I mean it completely could be. Um, but part of that has led us to. We released uh, uh, an iOS app earlier this year, uh, just at the end of. I think about the end of Q1. Hopefully we'll have an Android app within the very near future. Uh, that is a recognition that more and more artists and the creative process is happening on mobile. So I think there's a couple of different elements there. I think more people having access to mobile devices and more of these types of creative tools. And to be quite frank the quality of the hardware and you know, the speakers and the microphones within the hardware and cameras etc is enabling uh, content essentially music to be created on a phone. And I think when that happens we're seeing a bit of an explosion in the sense of, you know, people can make music who for one reason or another, maybe they wouldn't have thought to make music or maybe they just didn't have the tooling but the, the barrier to entry, um, both from a creative perspective but also from a distributional perspective with Distrokid, that's really, those two barriers are really falling down and I think those two things combined are ah, what's leading into the increase in content rather than something like generative AI at this moment in time. But I think it will, it will push the needle up, but I think it's going to take time. Mhm.

Speaker A: Based on what you say, uh, there, there should be more and more content produced by artists, uh, and even by non professionals. It would be easier and easier to generate that content. Uh, also you can do that possibly on, on your mobile phone. So does it mean that um, the outlook for music streaming is uh, is very positive, right, because this sector should increase and grow in the future?

Speaker B: I think that's right. Um, I think it's already been after kind of what some might call quite dark years in terms of the overall health of the recording music industry. Um, sort of post the Napster era, sort of to the early to mid 2010s there was a significant decline in the recorded music industry. That trend is really reversed and like things are going very well for the whole global recording music industry. Things are going very well for major record labels. Um, you know you can look at them, they're enjoying extremely good profit margins. Um, so things are going well for the top of the, of the music chain as well in terms of top artists. Um, I personally am a big believer that you know, if you want to create music and you want to be able to put it into the world and put it on these streaming services like you, you know, you should have that ability. Um, the streaming services are where people go to listen to music. You know, like a few years ago or maybe a bit further back it was still a little bit more disparate and certainly a lot of people still go to YouTube for example. Anybody can upload anything to YouTube. Um, I understand that there is an increasing amount of content going onto services like Spotify, but I think it's only for the best. I mean I think we see a lot of artists at Distrokid who are incredibly, um, successful. Um, but I don't think we're overly focused on that sort of demographic of artists. I think our service and our product is really about you can be anyone. If you have music and you think it's great and you want to put it out into the world, we want to make that happen for you. And I think if someone's passionate about music and they want to get out into the world, they should be able to. I think, uh, you know, we're moving in, we're definitely moving into a world where the relationship between uh, a creator of music or a creator of art and the people who can listen to that or consume that content, those barriers are kind of, some of them have been eroded, you know, through social media and like TikTok and people speak more directly to their fans. Uh, I think a lot of those types of artists that are kind of building their careers from the very ground up, um, like a service like ours is fantastic. I think they should have every right to be able to get their content onto Spotify and to get people to listen to it. And we do what we can to uh, empower artists across. Regardless of who you are.

Speaker A: Mhm. So we have been talking a lot about different use cases, how you can use data analytics, generative AI at Distrokit and also in um, music industry. But of course to implement those use cases you need to use some technology. And can you share us a few words about the tech stack and machine learning toolkit libraries that you're using at Distrokid?

Speaker B: Yeah, of course, I'd be really happy to do that. Um, so Distrokid, as I've mentioned a couple of times, has um, grown over time. It's been around for a while. So the stack has kind of evolved in different ways, um, regarding what we use specifically. So in terms of how we kind of manage a lot of our data and move things around, we're using a lot of AWS and then a lot of SQL infrastructure, whether that's, we have some MySQL we've also got, um, we use Redshift for a lot of our analytical processing. Uh, we use uh, an open source tool called Redash, uh, for our data, um, essentially for our kind of querying and dashboarding. Um, so that empowers you know, our analytical team, but also anybody in the company to go in and look at um, data queries, dashboards that we've put together to monitor different things. Um, regarding our sort of use of machine learning. We're definitely evolving here. So there's, there's more that we can do. Um, currently we do quite a lot of things just on EC2. Uh, but we are using a bunch of models. Uh, in particular we're looking at things like Whis. Well so we do use things like Whisper for audio transcription. Uh, we're using clip and um, AWS's recognition services to look at images and album art and make sure it's appropriate to be on the service. Um, but there's more interesting cool stuff we want to look at. I'm, I think one thing, you know, we've talked a lot about generative A.I. and I think one thing that generative A.I. requires is um, particularly in the kind of training stage is uh, that you kind of need access to this on demand compute for a very short period of time. And it can be quite spiky. Like you kind of need access to the compute and then you don't need it anymore for quite a long period of time. You kind of have your model artifact and then you deploy it. So we are looking at otherwise you know, different tools and things we can do that. I'm super interested in the growth of sort of infrastructure as a service or um, you know, cloud computer code. Um, I'm very uh, very interested in the tools that have popped up around generative AI. Things like replicate that or hugging, like hugging faces, um, infrastructure products that allow you to abstract away a lot of the managing of the infrastructure and just really pare it down to like the code itself. I think that's great because we're not a big company, we're not a big data science team. So anything that helps us abstract away managing of configuration and infrastructure is really interesting. Uh, I'm a big fan on a personal level of Eric Bernardson's modal startup that just had their Series A. I've used that personally. I think it's amazing the ability to just like plug some decorators into your code and access on demand cloud compute and then spin up you know, machine learning models or whatever it might be as a web hook really seamlessly. So I think that stuff is super cool and I want to, you know, I'm very keen that we begin to use a lot more of that. Um, but even at a more basic level some of the work we have to do is um, just um, in getting together better data pipelines and getting better single sources of truth about our data. Um, we, we're introducing more and more complexity into distrokid in the last two years as a result of our growth. You know, I mentioned the mobile app. Um, suddenly we no longer have like one place that we receive payments from. We now have two places and we now have two devices that we need to do uh, telemetry and usage statistics over. And you know, as you probably know, like going from sort of four to five platforms is usually not that big of a pain in terms of like evolve in your data stack. But when you go from one to two, you're suddenly like oh, here are all these things that I've written or got set up that don't necessarily scale. And so some of that is we are actually just figuring out how to do that a little bit better. Um, we're using the DBT core products um, to do some of that right now and it's working quite well. Um, but we definitely have more to do in this realm. And um. Yeah,

Speaker A: and I have the last final question because you have been working at a number of uh, music companies such as Spotify, where we both work together and also Eric Bernardson who you mentioned in the previous answer, uh, and also you worked at Utopia Music and now at ah, Distrokid. So you know uh, those companies and this industry very well. And can you share what, what makes uh, Distrokid different than other companies or, or kind of unique in your opinion?

Speaker B: Yeah, within the music space for certain. I think one thing I've really noticed at Distrokid, and uh, I'm slightly embarrassed to say it because I myself am not a musician as I said earlier but uh, so many people who work at Distrokids are musicians, like active musicians, practicing musicians come from a very musical background. You know, I think the, probably like the majority of people who work at Distrokids release content through Distrokids, you know. And uh, it's just, it's uh, it just creates a culture that is very focused around understanding the musician experience and trying to give something that is really useful for musicians and really um, really fair for them. We're making sort of uh, constantly making improvements and adding new features to, within the Distrokid product. But one thing that we've done a lot of is focusing on additional things that can help musicians like manage their, not just their kind of content but like um, their sort of careers and that like help them in their creative process without being too invasive. So um, we introduced the kind of mastering product that I mentioned earlier to master people's tracks. We have a video product that allows people to get videos onto YouTube and you know, realize the revenues from them. Uh, we recently um, completed an acquisition of a fantastic company called Banzoogle, uh, which is a long standing company, um, within the music industry that provides uh, of just an incredible amount of tooling for artists to create their own websites and to kind of manage their own um, sort of brand I guess you would call it. And that's just another way I think that we're trying to provide that kind of 360 experience to help musicians through at all points throughout their career. And I think it's great working at a company that is full of people who are musicians because they just, you know, they do do just really understand like this is something that is appropriate and good for musicians. And I think it means that we have a very good sort of moral compass about how we think about the decisions that we make, how we think about how we interact with musicians. You know, we have this 100% royalties gets passed through to artists. And that really comes from a place of really deeply caring about the musician experience. And I think we wouldn't have that to the same extent if we didn't have the type of people that we have at the company.

Speaker A: Mhm. And this is really, really great and really important. And this also concludes our podcast episode. So Chris, I would like to thank you very much for sharing your knowledge with us.

Speaker B: Thank you Adam. It's absolute pleasure to be here. Loved it.

Speaker A: If you are interested in getting notifications about future podcast episodes, please subscribe to Radiodata Podcast on Spotify, Apple or YouTube. If you are interested in being an expert guest in one of our episodes, please find me on LinkedIn and send me a message. My name is Adam Kava and I work at getindata, which is a data consultancy company. If you would like to learn more about our data analytics, AI, ML and cloud projects and our services, please visit us@uh, getting data.commhm.