Paul Starrett: 0:00

Ready to go. Hello, everybody. Welcome to another podcast by PrivacyLabs. My name is Paul Starrett. I am the founder of PrivacyLabs.. Remember, PrivacyLabs. is one word. And this podcast today is going to be with Dimitry Kushelevsky. And this is in a series of podcasts on privacy preservation, and democratization of data, which is the focus of this podcast and similar technology specifically, generally within the area of machine learning and artificial intelligence. Just a little bit of background on Dimitry and myself, we had the pleasure of meeting through an investment group about three months ago, we both are advisors in various capacities for a company called Ealax.com company that specializes synthetic data for financial crime. But since then Dimitry and I have had many conversations around this topic. And I thought it would be wonderful to tap his brain for this area in democratization since his company datasource.ai is specializes in that. And his background is really perfect for this topic. So we'll be talking with him about that. And I think without further ado, Dimitry, if you introduce yourself and your company, and then we'll just dive right in.

Dimitry Kushelevsky: 1:27

That's great. Well, well, thanks again, for involving me in your podcast. It's, it's an honor, and I am most happy to continue our conversation, which has been very productive and very engaging so far. So let's see. So where do I begin? So, as you mentioned, I am the CEO and co founder in datasource.ai, a startup that we were started with the sole purpose of democratizing AI, more specifically, data science in the form of machine learning, and making its incredible capabilities available to the entire world. Right now, it really is what I loosely describe as a 1% problem, 1% versus 99%. It seems that many people, many business organizations, many individuals in tech, are already very familiar with the concept of AI and what it can bring the specific benefits that it can bring, as far as improving their operations as far as bringing additional revenues and boosting their potential leads to boost their profits. The bottom line, if you will, however, very few companies out there really can boast that they've actually taken a serious strategic approach to deploying AI algorithms in their software infrastructure and their software stack. And, you know, it is, like I said, it's more of a 1% ai problem where a handful of the visionary companies with typically with big budgets, they're typically, you know, multinational global corporations, they realize that there is a great deal to be gained with very low potential risk at the same time. So they seemed perfectly comfortable spending some money on developing a data science team and making their you know, I should say, becoming an early adopter of AI when it comes to actual implementation of various AI algorithms, as well as data science tools overall in their in their operating infrastructure. Meanwhile, the mainstream of the business organizations out there are still very much left on the outside looking in. So far, if if a company wanted to deploy any serious AI capabilities in their software infrastructure, that pretty much by default, required that they hire and either hire an in house data science team, and acquire an actual infrastructure engineering team that would develop a physical as well as base software infrastructure to run data science and AI algorithms. And that of course, costs quite a bit of money. And it does require a considerable amount of expertise, which today is still in a great deal of deficit. It's still fairly hard to come by. And the schools of course, the power universities across the globe are producing data scientists as quickly as they can. But there is still a pretty significant deficit for that area for that specialization. So that, where does that leave the, basically the 99%, as I refer to them, so far, most of them simply have not been able to, to even seriously play around with AI and machine learning capabilities. And they've basically been doing what they've been doing for the last 20-30 years. Most of them, you know, who, who did want to, who did want to do some sort of a decision making implement some sort of a automated decision making in their software stack, they typically use rules based software, that is, of course, very limited, because it's not, it's not based on the dynamics of the immediate situation in the immediate scenario at hand. So to use a very common example, if you have an ecommerce store it, it, of course, can have some basic rules, rule sets, but built in baked into a script, that would, that would tell the machine or the controller to perform a certain task, whenever a visitor comes to, you know, looking for a specific recommendation, or looking for, you know, looking to do something in their store or purchase something in their store. That's great. But of course, that if you have a rules based algorithm, that's not based, that's not using AI, in essence, you're trying to trying to serve as this potential client by looking in the rearview mirror. And, of course, there's only so much that you can do, of course, you know, the really cool part of machine learning and AI, is that you can actually have a machine or an algorithm monitor all the real time details surrounding this particular visit, or in my fictitious example of an ecommerce store. And based on what it's seeing, it can make a real time decision that is a lot more likely going to result in in the in the purchase or in the customer being delighted, because he or she managed to get a great recommendation, when perhaps they least expected it. So anyway, the long story short, and that is, by deploying AI by by utilizing the toolkits that are available with machine learning data science, and other affiliated technologies in that space. Very few people today argue that there is nothing nothing to be gained. However, very at the same time, very few people, especially the smaller and medium sized businesses with typically tighter budgets and, and more limiting real human resources constraints, they're typically locked out, you know, just costs too much. And they just don't have that kind of those kinds of resources and expertise to, you know, to throw into data science or AI or machine learning. So that's basically where we come in, we are trying to bring the both the price points associated with AI and machine learning down to a point where a typical, you know, middle of the road, SMB business, should be able to afford it. And at the same time we are performing, we have implemented a number of unique features, such as automation, that would make it very easy for that type of a user that type of client to actually implement elements, functional elements of AI and machine learning in their infrastructure. Without that requirement that I mentioned before, without requiring that the they hire onboard data scientists or spend a lot of money on a data science infrastructure to complement their existing operational infrastructure. So that's, in essence, what we're trying to do and we're hoping that ultimately we can deliver a tidal wave of benefits to a very large number of of people and businesses that otherwise until now have been unable to, to access them.

Paul Starrett: 9:51

Great, no, I and that's a great lead in actually, I think you stated the the existing state where things are the 1%, and then the lockout, if you will, of the remaining 99%. And I think it'd be helpful to get down under the hood a bit more into what datasource.ai does. If listeners aren't familiar, there's a company called Kaggle, which was recently, I guess they were purchased by Google. And Kaggle, what they do is they put out a challenge or a problem, and they ask for people to submit to kaggle solutions. And if they are, if their solution is chosen, they're given a cash reward. Often that's, you know, 50,000, 100,000, it's quite a bit of money. But the idea is to get all of these contributors who are competing for that prize. And in so doing, they're getting really this very high quality very sort of, well, sort of the competition has drawn out the best of the, the those who are contributing what we call sort of the crowdsourcing. And what you're doing datasource.ai is taking the concept and making it much more available, kind of the Henry Ford, if you will, you're, you're allowing it to come to the masses. And so you have a smaller sometimes, you know, the cash prize, if you will, could be 5000, it could be free, really depends. But the idea is that this the SMB, the small to medium sized business, then has access, they put up a cash mount, like $5,000, I'm just picking names out of that are numbers out of a hat, you they then come to you, and then you get this competition. And I think that let me know if I haven't stated that properly, but also need you to state of I think you've got quite a few projects, going. how

Dimitry Kushelevsky: 11:45

We got it, we are definitely turning some heads and attracting, frankly, a lot of heavyweights in the data science community who, as we've already demonstrated, who are happy to contribute their skills and the energy and creativity to, you know, to help us become successful. Yeah, we've done a number of projects, as you mentioned, that, in essence, our data science competitions, but so far, or most of them, were did not have a cash prize associated with them, we just wanted to, you know, to try out our, our platform to make sure that the features and automation and other capabilities are working as, as planned. And at the same time, we wanted to test just the general assumption behind our business model, which is, you know, there is a very committed very high energy, very vibrant community supporting data science, as well as implementations of AI and machine learning in the mainstream businesses and other organizations. So, so far, we've been very, very pleased with what we observed, we are actually beginning to monetize our, our platform now. So it's very exciting time as well, because I want to to offer actual cash prizes, to, to the winners of the of the most successful algorithms that our contestants have got submitted. And also what we're doing, you know, thank you for bringing up Kaggle. While the concept behind crowdsourcing AI or machine learning algorithms is actually quite similar between what we do and what Kaggle does. But there are certainly a number of unique capabilities, starting from the differential between the markets, the target markets that they focus their offerings toward, versus what we're trying to do. So as I mentioned earlier, we're really looking to bring it down to both a very low price point, as well as a very low requirement of, of the expertise and other dedicated resources that a any given client would have to have on board in order to use our system. But in order to ultimately, you know, develop a high quality machine learning algorithm and implemented in the, in their software infrastructure. Typically Kaggle project still would require data scientists onboard those data scientists will typically come with the project, you know, the customer, the client would be expected to bring him in the cash prizes with Kaggle are significantly greater, I'd say typically on the order of magnitude greater versus our target cash prize values. So by doing so, once again, we're trying to really bring all these great benefits of AI and machine learning and data science into the global mainstream. So obviously, we're you know that that entails that we would try to turn it into a very much a high volume, low barrier to entry type business model, and want to have lots and lots of businesses, you know, who could, who could, you know, realize very quickly that, hey, I can actually for for very little money. And without having to go and hire dedicated data scientists, to my team, I can actually go and develop one or more machine learning algorithms that are going to be high quality, they're going to be designed by humans, by expert humans. And they are extremely likely based on that indicators that we've seen from the earlier deployments, they're extremely likely to improve our business and grow our bottom line, which is ultimately what we're trying to do. I mean, you know, ultimately, as far as our purpose goes, that we behind our company, behind both of our, but myself and my co founder, Daniel, we are really trying to, you know, we are passionate, obviously, we're passionate about AI and data science and machine learning. And we are really focused on bringing all those great capabilities, all those great, fairly easily attainable benefits that the, you know, that customers can utilize, right down to the average business, the average organization around the globe, no matter what their budget, no matter what their size, no matter what their, what their ability is to, you know, to hire on board expertise and other resources. So that's obviously because of that deep desire that Daniel and I share, and have shared from the very beginning, we have developed and launched a platform that is highly automated already. Although, of course, without question, as we progress as we grow. And we have additional developer resources, of course, we're going to continue to, to enhance it. And, you know, and to add additional features and capabilities that that are only planning today. And the ultimate benefit is that as we get more and more clients, utilizing our platform to crowdsource high value, high capability, high quality machine learning algorithms, as they deploy those algorithms, they will undoubtedly be getting very impressive results based on everything we've seen in all the studies we've read so far, they are really setting themselves up for a great deal of additional success, even if they're a successful company already. So that, of course, is why Daniel and I are very excited to be doing what we're doing. And we're even more. So more. So we're even more excited about the future that, you know, that this technology holds that we could potentially bring to the mainstream business customers around the globe as we grow as a company.

Paul Starrett: 18:22

Yes, and that's, that's great. I it it let's it leads me to think of the the crowdsourcing, it's not only does the individual company get the benefit of the, of your platform and your expertise between you and your co founder, in addition to all of the teams that are competing, to satisfy some goal that the competition so to speak, is put to, there's also, this is going to lead into, I think the part here where we're gonna get into the challenges that come with this, that what you can do is you can have, let's say different companies that are perhaps in the same vertical the same domain, share your information, to gain the synergy across their different insights, learn from the machine learning efforts. The problem is, especially in highly regulated industries, with if getting the data is the big problem. And the one of the biggest barriers there, of course, is privacy regulation and data protection laws. And the idea there is that there are techniques, there are a solutions that allow you to essentially create a different data set that's called there's various things here now, it's a big, it's a fairly large topic. We cover this, I just finished a podcast with Patricia Thaine, which you'll find on our website which discusses privacy preservation technologies in the grand scheme. But for right now here with machine learning, we're going to focus on synthetic data. What that is, is is a method by which an algorithm will take the original data that contains private sensitive data. And it replicates it. But it leaves behind any remnants of the sensitivity, or of the privacy of the underlying data, thereby kind of lifting it up and out of those concerns. So now you can share it it's not a panacea, there's a thing called the privacy budget, which says that the more that you remove the privacy and sensitive information, the less valuable your data becomes to a machine or machine learning algorithm. And it's not a it's not a simple process, but it's very doable. And so Dimitry, I think, you know, Ealax company mentioned earlier, they do this, and be able to do it for things like a banking and financial services. And I know, Dimitry, you personally have quite a, quite a bit of background in this area of financial services. What is your perspective on the promise of synthetic data and your thoughts on what it is and, and, and how we expect to see that utilized not only for a company to do it just for the internal purposes, but then perhaps to share it with other?

Dimitry Kushelevsky: 21:12

Yeah, absolutely. So without question financial, the financial vertical financial industry is one of the one of the verticals, that is really, really well positioned to take advantage of AI and the power of the capabilities that that it can bring to them, again, with, you know, with the help of a company like ours, for a very low cost and a very low resource requirement. And, again, it seems that, I guess, because the financial industry is so close to business, and so, so close to recognizing the the material aspect of what this kind of technology these technologies can bring, they they're getting it, you know, they clearly they're, they're sensing that this is not just a fad, AI is here to stay. And, again, there's they're seeing like the smaller local institutions are seeing that the the larger brands in their industry are deploying AI either I would say the, you know, the larger financial vertical representatives are among those early adopters who, you know, who have done some strategic early deployments, and they actually have benefited from them pretty significantly. So, you know, what, what does what does does the future hold or what does what kind of capabilities, what kind of benefits does does it hold for for Finance? Well, there are so many great applications, right, I normally start looking at any business opportunity or even a use case scenario by by examining the what what the customer's needs are, and in this case, in the financial vertical, the customer's needs are quite extensive, right, they are the most of the banking institutions and financial institutions already have considerable amount of data that they have been collecting about their customers just as a part of their day to day operations. And of course, because they are required to do so by law, right. So, for one, they already have a great important ingredient that many representatives of other verticals may or may not always have. So, they have the data, they also have very specific means such as they want to remain competitive, they want to, they want to be able to offer new services, they want to target their, their marketing and other customer focused materials better. And ultimately, they of course, they want to save on their operations as well. Another another huge opportunity for the financial industry across the board, of course, is something that we discussed earlier. Is, is the fraud and, you know, criminal activity prevention. So AI, of course, I'm you know, I'm I'm very excited, you know, banner waving, waving, you know, person in the AI ecosystem. So yes, I do admit that I might be a bit biased here. But AI, I really would, would strongly submit that AI provides a tremendous opportunity, perhaps much more powerful than any other source of tools available today, to address all of these use case scenarios, and they're really exciting part to me here is that we would be, by developing AI algorithms and other AI based solutions, we could directly and very positively impact you know, those customers and meet their needs. You know. So that's, that's really exciting part, ultimately, everything has to, you know, begin and end with the customer. So anytime that we have, we have a customer who already has a demonstrated set of needs that can directly impact their, their business in a very positive manner. Of course, any business person will be very excited to offer their platform or their solution to help their their users get and get exactly that effect. So, yeah, there's a, there's a lot, a lot to do a lot of opportunity. But of course, there is always, as always, there is a challenge. And the challenge is quite significant in financial spaces, that has to do with regulation. And it has to do with the severe privacy protection regulations that virtually all the financial institutions have to abide by across the globe. Right. So that is one big challenge that that without, with that, unless we find a way to solve it as an industry, I think, you know, Ai, and machine learning and data science will be extremely limited in terms of the depth and breadth of those benefits that we can deliver. So having having companies like Ealax around producing very close proxies for the customer's actual original data, however, without disclosing any of the any of the private or personal or confidential information associated with the bank, or its customers, or without with institutional risk customers, could may very well be the difference between all those institutions, being able to take advantage of these great, but your business benefits and not being able to do so. So it's really quite a big development.

Paul Starrett: 27:20

Yes, I agree. And I think I wanted to sort of slip in an elevator pitch that I have to kind of encapsulate what you said about, you know, how data is becoming much more vexing even for the midsize, and small companies. Because, as we know, the the amount of data that companies generate is growing exponentially every year. And the only way to really wrangle it is with with machine learning. That's all you're left with. So it becomes the new normal becomes the best practice. I think some unique things that we can share with our listeners, is that synthetic data does allow not only for us to drop out the sensitive or private information, again, though, want to emphasize it's not a panacea. There's there's some, some knobs to turn. And there was some loss of insight, but often no free lunch, right? Exactly, exactly. So privacy budgets, you got to pay somewhere. But I think generally it's very much a net gain. But there's an upside to that as well as it with synthetic data, you can actually gain more insights from the underlying data that go above and beyond what you'd expect to build in a machine learning model from that data. Because the synthetic data can generate new types of transactions and new types of scenarios that a machine learning algorithm can then use. It also has the ability to some other issues around regulation has to do with explainability of machine learning, how's it working? Do we know what the model the machine learning model is doing? You can you can add into this synthetic data, metrics, and other information that help you establish, you know, how the explainability, which is a very big piece of the privacy, regulations and so forth. GDPR has specific requirements around that, as do most laws, and just for just a picture of my own, you know, blow my own horn here and pay some bills, that's what PrivacyLabs, does we help come in and make sure that I have a background in machine learning abd law. And so I'm able to help bring things together, get the machine the get the explainability in there, and to make sure that the the compliance professionals understand the technology, and what's happening and make sure that all kind of comes together, profitable and compliant way. So that's kind of our role in this. And I of course, look forward to working with you and, Ealax and other companies to to sort of bring this to the market. I think that's, I think from the standpoint of the so that really the goal here is that democratization of data and I think maybe we can finish on this topic. That we've basically covered the idea that the individual institution, whether they're small or midsize, really, I think is where the the, the, the issue of the need is, is most vexing. The data is getting bigger and faster, more complex. And then machine learning really is the best way to save money and reduce risk and so forth. But this also the ability to build to make a better world and Dimitry this is a big piece of Absolutely, it's in your heart is that, again, could we have, let's say financial services, institutions share all of their data together to build kind of a, for example, a fraud machine learning model, that is sort of a superset of all of the intelligence has come from all of the things. Again, I think that when we get into things like synthetic data and other things, that becomes much more realistic. And you have this sort of crowdsourcing in its own right, in that regard.

Dimitry Kushelevsky: 30:57

And you get to use the wisdom of the crowd to solve solve some of the biggest challenges that were dogging the entire industry across the globe. So yeah, this is one of the many excellent value points behind the entire technology.

Paul Starrett: 31:14

Yes, yes. In the area of for those who are a little bit more maybe savvy in the direction of data science, a thing called transfer learning where you're taking, essentially, the typical case is deep learning neural networks, and you're able to take the prior models that have been built, and then leverage that background. Transformers are a typical example. But again, that's that just sort of a aside, mentioned, for those of us who are a little bit more into data science. I think that kind of rounds up again, the purpose that the idea here was the democratization of data sharing, it's being able to leverage democratization for crowdsourcing of information around a specific problem for a company, such that they can then become the can enter the market and remain competitive by being able to leverage and have access to machine learning, but also in the ability to have domain share information for the common good. So I think that we've done a great job, frankly, I think in this what is roughly half an hour,

Dimitry Kushelevsky: 32:24

there's a lot of ground to cover. For some, like yourself, I'm sure you know it, there's a great temptation to get into the weeds, because there are so many great use cases and so many great applications, and ultimately, so many incredible benefits, business and personal benefits that we can deliver to literally billions of people out there with this with this type of technology. That, of course is very, very exciting. And, you know, frankly, that's, I think, very much a part of our future. You know, if I just read a PwC sourced study recently, where they claim that by the year 2030, they explore we expect that AI is going to add a little over $15 trillion, that 15 trillion. Yeah, one 515 trillion dollars to the global economy. It's incredible, just absolutely incredible, frankly, even today, closer to home, so to speak, or closer to our timeframe, right now, the machine learning but your industry is measured somewhere around nine or between nine and $10 billion. Obviously COVID kind of played with those numbers, like with any other numbers, but I believe that's still more or less where we are today. But the really exciting news, and I believe the study this study, mistaken came out of McKinsey, they are actually forecasting a 39% year over year, compound growth rate for the next foreseeable future, I believe the by the year 26 or 27, they're expecting this number to go grow up to around 120 127 billion. So it I mean, these are astronomical numbers. You know, and you mentioned earlier that, yes, there are certainly multiple applications that are multiple entrants into the AI and machine learning sourcing space. And I'm, I'm certain there will be more I don't think it's that big a reach to to forecast that it's going to get better and better and bigger and more. You know, that densely populated as far as the AI industry goes. But my you know, the way I see it is there's so much great potential, it's truly just an ideal, you know, textbook case of plenty for mentality, it's something that we are going to, we can, we can build new solutions within to develop a tremendous amount of value added to, you know, to literally millions, if not billions of customers. So there's plenty of, there's plenty of good to be done, you know, that's a really, really exciting part. For everyone who is already in this space or is considering, you know, entering it, including the folks who are potentially going to be our future customers, we welcome them to come and check us out. And, and, you know, we offer a free consultation for anybody who's interested in exploring what, you know, what we offer, and how it may be able to benefit their their business, their operations or, you know, overcome any other challenges that they might be facing?

Paul Starrett: 35:59

Yes, yes. And I did want to sneak in here one more comment about an and then I'm gonna ask you to retreat for your, your closing thoughts on what you think we haven't haven't covered or something you think needs to be emphasized. But I think one of the other things that we keep talking about synthetic data. And I just want to iterate the reason we say that is because Gartner has predicted this 60% of machine learning will be based on synthetic data by 2024. That's right around the corner. So I think that kind of gives us a sense of, there's an there's an area, and I'll make this brief because it's a technical area, that the software development lifecycle has really moved to what they call an agile framework, which requires very quick turnaround. And that is the new normal for the development of anything, any kind of software or any solution that's being used by enterprise. And the problem is, is that to get the data, it takes a long time, contracts and laws and other things require months. And you don't have that time when you have an agile process in software development that requires a daily kind of turnaround. So this synthetic data allows you to generate that data much more quickly and get get to pay dirt. I just wanted to do that. That's a very new hot topic that we've kind of tripped over here from other discussions. So other than that, I'm going to finish here. I will anything, Dimitry, you think we should, you know, we've got a few minutes here. Anything you think that we should know, that we haven't discussed or anything you want to emphasize?

Dimitry Kushelevsky: 37:24

Yeah, well, the one of the most interesting challenges that we are up against right now is, is we rather obviously, don't want to boil the ocean, if, if you if you know what I mean, there are so many great use case scenarios, there are so many great applications for AI for machine learning for, you know, quite literally running data science competition, that we you know, we have to be very judicious as far as which ones we pursue, it was a great temptation between both founders to try and just go after every interesting opportunity, every challenge that has a real business need and real data behind it, that the customer may already have a potential customer. But we find ourselves deliberately, you know, keeping ourselves disciplined in a way that we want, you know, we're trying to validate our major assumptions that will rather obviously, you know, provide us the, our go to market and our business, evolution projectory for, you know, for the foreseeable future. So, I With that in mind, so yeah, it's a great problem to have. And with that in mind, I, again, I want to welcome any, anybody who's interested in playing in this space, and even just checking us out and seeing and discussing with one of our experts, or one of us directly, what we can, what we can do and how, in specific terms AI and machine learning can, can help them overcome their challenges and grow their business and bolster their bottom line or take better care of their customers. So once again, I of course, we would love to, I would love to welcome additional people who are either as excited about AI as we are, or perhaps they're just intrigued. And they, you know, if nothing else, they want to see, hey, let's talk and let's see what what this technology and technology may potentially have in store for them and their business. SoI again, I welcome people to listening to this or intrigued about the potential benefits that they can gain with AI Data Science and machine learning, I welcome them to come visit us. If you know today's if they are interested, they, if they are intrigued by what you and I just discussed, they're intrigued by the content that we've posted on our web page. I, of course, would love to chat with them, and they can just click on the free consultation by and schedule a few minutes to chat with us, I think, you know, every single conversation is, is very interesting to us. Because, again, it kind of helps us to triangulate the most promising opportunities for us to deliver maximum value. So don't wind up boiling the ocean, but we ultimately wind up, you know, meeting the meeting our our mission requirements of our mission and helping businesses accomplish their Akash accomplish their goals for success. And hopefully better than any other alternatives out there in the marketplace, which I do strongly believe that we can. So thank you for thank you for the opportunity.

Paul Starrett: 41:10

Yes, no, my pleasure. And I just so people know, I guess the website is its datasource.ai. And it's all one word, no hyphens, no dots or anything data source.ai. And I believe is it dimitry@datasource.ai? day?

Dimitry Kushelevsky: 41:25

Yeah, dimitry@datasource.ai. You know, if and that's, believe me, just having my first name is a blessing, as you know, because in this email address, because I have a long, you know, Ukrainian last name that that would confuse anybody. So, yes, but I, of course, would welcome you know, any, anyone who wants to reach out and, and connect with me directly.

Paul Starrett: 41:58

Great or they can go to your website, as you indicated. Great. Well, listen, I'm just going to close out here with some thoughts on PrivacyLabs sort of role in this is that the process of bringing artificial intelligence or machine learning into your enterprise infrastructure in one form or another, is a horizontally kind of active topic. And that's where we can help to look at the security requirements, the compliance, I have an attorney who's kind of specialized in compliance law, I'm much more technical, but I can help discuss the topics with the compliance folks and help sort of scope things and one thing we do in privacy Labs is we are we work with partner companies like One Trust and BigID, and TrustArc, and at one another, one of my favorites is Centrl. That we can use those tools to help kind of herd the cats to kind of bring everything together. We specialize in machine learning and automation and an audit so that we can make sure that everything's going the way it would be expected either by by way of a regulator or to to make sure you're, you're covered legally at some level. So that's kind of what we do. And again, Dimitriy thank you so much. And I think we'll close out here, and I'm sure

Dimitry Kushelevsky: 43:21

I wanted to give you a quick plug, Paul, yes, because I deeply appreciate what you do. As far as opening the gains for potentially a very large number of, of business owners and business executives, who, because of you and your work, will be able to take advantage of what we offer. So that's I really appreciate having having met you and having had a bunch of really productive conversations that we had already. And I look forward to continuing very much along the same lines.

Paul Starrett: 44:04

Thank you. Those are kind words, and I wouldn't disagree with you if I say so myself. I think we've really we've really positioned ourselves and it's usually with with my guidance directly, personally. Yes, we're sort of the concierge if you will, to kind of help people get in and cover all the bases horizontally and peripherally. So great. With that said, we will close ourselves out here. And Dimitry, we will have another podcast soon. Probably one of the updates or some other vertical or something. But thank you again. And thank you listeners. I hope that you learned a lot and watch for future podcasts from us. Thanks. Thanks all.