Greg Irwin 0:19

Putting our discussion group with BWG Dell and TigerGraph, we're talking basically leveraging graph for, you know, customer analytics, check me out anti money laundering and fraud for financial services. Tremendous group here today. So this is going to be a lot of fun. Let me give a little bit of the overview of how we do things and what y'all can expect today. And I'm looking for a fantastic forum. First off, we'll spend just a minute or two here just with with intros, and Daniel and Michael guys are, are going to be are going to be co hosting with me today. So thanks in advance. Um, the more interactive This is, the better. We have this chat window, I recommend everybody open it up. And I'm going to ask you a question or two through our session and do us a favor, you can ask each other questions. If somebody says something and you want a little bit more info, drop your question or comment into the chat. And I'll ask the group to try and interactively here, use that as a back channel or a side channel to basically provide all the info y'all need to make sure that this is useful and productive a meeting as can be. Lastly, as you know, we're here on Zoom. So if you're able to turn your camera on, please do it helps a little bit with that engagement and a little bit more of that, that human touch. And with that, let's get going. It Michael, do you want to go? Go next here and give a little intro for your focus over at Dell?

Michael Krogh 2:07

You bet. You bet. Hello, everybody. My name is Michael Krogh. As mentioned, I'm with Dell Technologies. I am part of our AI and data analytics practice. My role at Dell very specifically is to work on strategy or develop strategy for our go to market messaging, how we engage with partners, how we develop our messaging out to the world with regards to some of our solutions and products. TigerGraph, obviously being one of those that we're going to talk about today with some careful criteria. So I'll talk a little bit about that here in a second.

Greg Irwin 2:43

Beautiful art Daniel you're up, please give an intro and a little bit on I think people know know what Dell is made may not be as clear on what TigerGraph is. So spend a minute as well and just explain what TigerGraph is.

Daniel Eaton 2:57

Yeah, thanks, Greg. So TigerGraph is the leader in graph technologies. We are a startup that created a new type of graph and analytics platform. And today, we're used by eight of the top 10 banks assets under management for everything from call center use cases to fraud in financial crimes detection, for AML, payments, fraud and KYC. Trade risk, and, and some other use cases. And I'm responsible for business development, with a number of our strategic partners, Dell being one of our best partners. And we're super excited about just having an interactive discussion today. I'm located in Minneapolis, Minnesota, and love, love traveling to New York City and in the Bay Area and other places around this great country. So nice to meet everyone virtually.

Greg Irwin 3:55

Before we get into it. Just one note. This, I don't like I grew up with about 20 years in nit, where I did a lot of meetings. I really don't love the one person presents Everybody listen, it's good. We're gonna do some of that, because there's some information to share here. But when you look across the Hollywood Squares that we've got going, it's really pretty phenomenal in terms of peers you have across the country at different institutions. And at the end of this call, we're going to send out a list of everybody's name. So as you go through this call, I'm going to ask everybody to basically set a personal goal for yourself, connect with one person across the grid. It doesn't have to be TigerGraph, or Dell or BWG. But as we go try and find one person across this grid to connect with. And, and with that, let me bring it back to Daniel. Daniel. Let's talk a little bit about the use cases. Let's get started immediately here. With how people are using graph databases and TigerGraph to help with know your customer, and AML. Okay, we have audio just check there up that audio there.

Michael Krogh 5:15

Yeah, certainly. So the sum of the use cases really span as I mentioned, a number from credit card fraud looking at is an applicant connected to potential fraudsters merchant analytics, looking at communities or clusters impacted by fraud rings, credit scoring, trade surveillance, crypto forensics, and, and some of the really relevant use cases around ransomware. And how you can use crypto forensics to uncover that. But let's let's zoom in a little bit on the end to end AML solutions that we've we've built. So TigerGraph is a graph database and analytics platform, and we're a Gartner and Forrester industry leader. So we have the highest performance, most scalable graph technology platform for the enterprise and why people use graph, in addition to all their other kind of cornucopia of big data technologies, and AML and KYC technologies, is because we give the customers the ability to see connected data in context. So what what that means is in graph, relationships between data are preserved as first class citizens that you can explore very, very easily and do things that traditional relational databases and no SQL databases simply can't handle. So to zoom in on the ML use case a little bit further, I've one of our customers was looking to improve its networking and link analysis capability for anti money laundering in three ways. They had connections between open work items and situations of interest that needed to be identified and available for analysts and investigators, they also needed to be able to conduct a very comprehensive and thorough ad hoc review of an entity should it display the connections from an ecosystem around that that point of investigation. And then thirdly, the system should be able to identify connections and situations of interest, that lead to productive investigations to inform the creation or hibernation or escalation of work items. And so as you can hear all that stuff, even for those who are on the data side, and not so much focused on AML, and KYC, you can see how connected that that those types of activities are work, you want to be able to click on something of interest and and see how that that transaction, that person or entity are connected to all sorts of other persons and entities and transactions. And then you want to be able to explore that in real time and apply analytics and advanced machine learning and graph algorithms to understand hidden relationships in that in that data.

Greg Irwin 8:31

Daniel, and you want to go through a single case study? I know I think we have Scott on from Expero. I think you all saw yeah. Oh, yeah, real, maybe Yeah, start early here with to kind of kind of drive

Daniel Eaton 8:48

I think, I think that'll be an awesome way to just throw a little grenade in the middle of this frozen pond to break the ice. So. So Scott's actually got a really quick, really quick demonstration of the technology. So this is what's actually this toolkit is a toolkit built on top of TigerGraph. And so customers can can use this toolkit. The toolkit can also be illustrative of what's possible. So if, if financial services institutions have a preference in building their own or have some type of interest in, in making something better, they had the ability to do that, but Scott, you want to maybe just hop into the hop into the demo real quick.

Scott 9:37

If somebody could let me share, that'd be awesome.

Greg Irwin 9:40

I think you might be able to hear me say I'm going to try and make you that make you host and maybe that'll do it.

Scott 9:46

So I've got a fun warm up exercise and this is this is one that I think will get a solid wood frame of mind. So interesting view here and this is anecdotal with tongue in cheek so for those of you that are not familiar With graph or graph analytics, does anybody know the difference between Katy Perry and Barack Obama? Long, long, awkward question. Anybody out there? Anybody that knows graph somebody from the field? Maybe?

Greg Irwin 10:15

Let's put a jump in with your comments. I like the idea of or put your comment into that chat.

Scott 10:26

It's not a trick question. But what's the difference? Yes, Katy Perry has purple hair. Micah you have a good, you're going in the right direction. So one is on a pet blast, and the other is not? What do you see about the bubbles around Katy and Barack Obama, you start to see something called cohorts or connections of data. And where we're leading in this discussion now is the concept that Katy Perry is as popular, if not more popular than Barack Obama. But what you start to see is singular connections. I may follow Katy Perry, on Twitter, I may see her events. But I am a single bubble. I don't know anyone else that is connected. For some reason, Barack Obama has more centrality or an algorithm that we can run, which shows that he has communities of more influential people. So for instance, people who follow Barack Obama in a graph will show us that he has other ex presidents and those people have connections of other influence or other things. What we're starting to see now is the data inside of a graph database mimics this same solution. What we also see now is, let's move to fraud. What's the difference between an online person who helps us scheduled travel and a coyote or human trafficking network? And the answer is inside of a graph database. It's the connectivity, meaning lots of unrelated people will books and scheduled travel, and inside of connections for human trafficking, you have coyotes that have dependent network networks. So what we're going to do now is we're going to use the power of this graph data structure to really sort of step into that now. What does that look like? Well, in this sort of quick investigation, we have something called a SAR. And if you're if you're familiar with what a SAR is out in the fraud world, it is a suspicious activity report, which is a very official thing that we have to file with fincen and other folks. And so we can look at people that are on these lists. But that's exactly what I'm doing. Now. I'm looking at this SAR, and I can see other kinds of information, I can click on it. And I can say, well show me where those things occurred. Right. So as Dan mentioned, right, we can see that this transaction, it may be a wire could be a phone connection, could be a you know, inbound or outbound, etc, right, we're going to use that same metaphor we just talked about with Barack and Katy, over on the right, what we start to see now is I can see payments, and I can see other kinds of things. Well, as I double click into that, now I'm using this power of the graph. And I can start to see what's going on, I can see some capabilities to move that I can take my map out of the way. And what I want to do now is the investigator wants to look into this concept of this connected data. The person Parker, in this case here works for a company, there are also things that happen, but I want to now do is look at things like risk. And what I can do now, along the bottom is as I move things through time, I can recreate well, Parker got a payment, he used his phone, He then made a debit, and so on and so forth. And so part of that is I can start to zoom in and I can say Where are these different elements? What are those things? That what I'm doing as an investigator over time now is I'm double clicking and saying, well, maybe I wonder if Nolan farmer is related? Well, that's interesting. Nolan and Parker work for the same company. Right. And there may be other connections. But the sort of quick gist of this is, in this connected data. I don't know where I'm going. So the interesting thing about a graph database is that I'm asking the graph database, who is connected? Where did they work? What other people were there other persons were there, other companies are there, their emails, transactions, and so on and so forth. And this allows an incredible amount of real time interrogation that is very, very, very, very difficult to do inside of a traditional data structure and That's really sort of the big epiphany is we can ask a graph a question, we do a very similar approach in customer 360. And so it's a little pause there. That's really sort of, you know, to get the juices flowing. That's what we're trying to do now. Right? So some of these challenges I see over here on the right. It's like somebody gave us a good one here is, you know, this fincen, recording beneficial owners and customers that are legal entities, but what you just saw, right is this concept of now the ability of how we can ingest data, and I'll share one more time. So there is this ability now to look at over here. Here's our fincen data. Over here on the right, I can look at open corporates, which tells me does a corporate structure exist? And so what happens now is in the ingest of the data, brought to us by TigerGraph and Xilinx on a Dell machine is I can do that in real time. And I can look at who exists who's applying, are there? You know, do you have existing systems with work or other kinds of things, I can look at your career trades, right? So I can look at everything from an account to a wire to a trade in real time, and then I'm able to create that risk scoring, which is what I just showed you in a few seconds ago was what does that look like? So that's a great question. In the box, any questions? I mean, this is a lot going on, right? I mean, the whole point of this is to find these features, to have our large banking customers, you know, they lowered the false positives by roughly 15%. What does that mean? That was roughly $20 million in one quarter, where they were able to find that they were able to streamline their users to go focus on those kinds of things, etc. So we talked about data and data layers, we can look at all kinds of data, whether it's address, social security number, the other thing that we can do is secure it based on your user or your user group. So if you're an investigator, in AML, you don't have to share what cyber is looking at maybe when you pull out a thread like I showed in that diagram, it does alert you that you need to coordinate Well, with TigerGraph, you can actually open that up now. And it's on a need to know basis. But now the cyber team needs to know about this individual case. And we can do things like that. So it's very powerful. It's very flexible. And it's a different paradigm. Right. So it's not just sort of a static Tableau version, where, you know, we'll tell you what happened yesterday, we can see things in real time. And the other construct that Daniel mentioned was, we can run machine learning, which means that I can try and predict I've seen patterns yesterday, where are those potentially going to be today? Where are they going to be tomorrow? That's another huge aspect of what a graph database specifically TigerGraph is really good at that speed on Dell technology.

Greg Irwin 17:57

I have a question, I'd welcome others to raise a hand, there's something called reactions here on Zoom, you can just put a raise your hand and I'll pull people in. Scott, I'm gonna get started with one. One question is, you're pulling together a lot of information in a sophisticated way I can I understand how a large financial institution might be able to do it. But what about a mid size a tier two, tier three, or a credit union that may not have as large of a risk team? You know, what kind of resources does it take to put this together and to operationalize it?

Scott 18:35

Well, the beauty of it is that the that that pictogram I showed you previously, inside of that was this engine, right. Think of it as a race car engine, right. It's a Indy 500 race car engine. Well, what we've done in that last user interface, and we've got, obviously more, you know, don't want to turn this into a sales call. But we've got more screens that are designed for specific users. So if I'm in a credit union, and I'm a level one investigator, it literally says fraud, or no fraud, it alerts them in this sort of pattern to say, this is something you should care about human. And it's very easy to do. The second level that we showed in that previous screen was somebody who can pull on the string. But again, you saw no code. So it's very easy. And it's actually designed for those midsize businesses where the business folks can go about their business. It's fast, it's intuitive, it's flexible, and it's easy. So that there isn't a large burden on that now that in the race car engine underneath it is is sort of that satisfaction knowing that with Dell technology and the TigerGraph technology, it's built for your business, if you need to grow to be the biggest or the second biggest or however big you want to grow or multinational. That's the other thing we see a lot of, it's designed for that should you need to grow and scale. And that's sort of the beauty of it is it's simple, it's easy, it's fast, we can turn it on as as fast as you to eight weeks, some of our bigger customers are 10 to 12 weeks, but still, that's not a long time. And so, you know, direct question. And answer is, it's fast. And it's easy.

Daniel Eaton 20:10

Yeah. And and to elaborate a little bit on that one also, we find with, with midsize and smaller financial organizations that they have an existing technology, many times it's from a turnkey end to end vendor like, nice atomize or, or actually, sometimes a mix of multiple vendors. And the beauty is, when we provide that, that X ray vision on the connected data and we have these financial crimes, specific user interfaces, we can bolt in right on top of adjacent to your existing solutions, and complement those, we don't necessarily need to be rip and replace. It's many of our financial services customers, we're in addition to, to give X ray vision into that connected data and those relationships and, and we typically also find that the system provides a payback and in much less than one year, in terms of in terms of financial benefit, not to mention, in this specific set of use cases, we think that there's there's also non financial benefits, just reducing the risk around legal penalties and the regulatory side of things. So we're super excited about this use case. And and Scott kind of hit the nail on the head when he said, we can turn around a proof of concept in in six to eight weeks and actually demonstrate this working for a bank using using their own data. And that's that's actually true, both of the very largest financial organizations. And we've we've done that and a number of the top 10 banks assets under management, as well as medium size and smaller financial institutions.

Greg Irwin 22:08

Guys, thanks a lot, let's let's do this, let's, let's switch gears just a little bit, what what I enjoy about these sessions is, you know, kind of mixing, you know, kind of a solution like this with just what other people are dealing with on a, you know, kind of a real time basis. So what I'd like to do is go across some financial institutions and talk about some of the operational challenges or initiatives that they've got. And what we always find is somebody else across the group might have an answer might have a, you know, a common challenge. Or, you know, or, you know, can set up an additional conversation. So, I'm going to do that a little bit here. I'd like to invite in Linda. Linda, you're up on the grid here, just to my right, so nice to meet you. Thanks for thanks for joining, do us a favor, just give a real, real quick intro of what your focus is.

Linda 23:08

Thank you. And my thank you for the introduction and happy to be here. So I have been in AML, actually, since 2002. And I have seen this continue to be I'm going to say headache for financial institutions to stay abreast as to how do we deal with what our regulatory requirements are. And compared to, you know, tech, not of technology is great, but so much data. And how do we I mean, I was quite interested when you mentioned that, you know, having the 15% drop in false positives, not as significant for financial institutions. But it's also looking at and you know, we don't have a crystal ball to sit there and say, what's gonna happen in the future, and our regulators when I've seen around the world, I previously worked at HSBC as the group head of AML policy. And part of that was looking at all the rules and regulations around the globe. And one thing I saw very much in common was the expectations of regulators for financial institutions to protect their economy, do not let money launderers and do not let terrorist financing in and expecting more and more. And it does get worrisome to see you know, when you have new things such as heat the US in the new definition of a domestic terrorist, you have challenges with ESG. Now climate and looking at this and looking at the technology that we have to use is trying to say, what can we find? And also we still need to balance that we have customers, you know, these are customers that you have a business trying to bring on newer customers. They want to ask less information, and I'm looking at this saying okay, this Would be great in a periodic review, potentially saying was my customer acting as I expect them to do? But what a how, what do I do about how do I know who that customer is? And how can I use this type of information? So we're faced with quite a bit as financial institutions. And I do feel I have to say, more for the smaller size financial institutions, because the burdens are just as great or even greater, because it has to be done with less people.

Greg Irwin 25:26

What's the What's one thing that you're working on? Because you know, this is a journey, you're not going to solve? There's no salt. This is just gradual, constant improvement. So what's over kind of a 12 month time frame? What's one big initiative? Maybe Maybe it's reduce false positives? Maybe it's better reporting? What's, what's one big bucket that you and your team are focused on?

Linda 25:51

It was it's really the Periodic Review, how do you come up with a sustainable process and maybe not having a set one year three year five year review, but having the ability to only look at that customer, when something drastically would change? And how do you prove that to the regulators, I don't have to go in and do a tick box routine to say is this customer who I believe he is on a periodic basis, when you think of your customer base, most, many customers don't change. And the cost that goes into performing a periodic review for a financial institution is really it, you know, beyond what we should be paying? Where we can be spending that time and effort on something that means more? So I would say it for me, it's been the Periodic Review, and saying, How can we have a sustainable, consistent review that I do not have to go in and physically look at each customer? Right.

Daniel Eaton 26:52

And graph, that's a that's an awesome use case for for graph technologies. Because what we can do is take in at the world's largest healthcare company, we take in 350 different data sources, and have a unified view of that customer, so that you can truly understand Are there any areas that are out of out of your expectation or, or are worth further investigation and drill down. And that's, that's a very common challenge. That's, that's great. Thanks for sharing that.

Scott 27:29

Well, the concept that we you know, when we got into this webinar, one of the things that that graph is really good at, which believe it or not, is an identical use case, it's called customer 360. or know your customer is literally inside of the data, the way the graph database sees it, it's the same use case. What that means, then is if I have a customer journey, and I get a wealth management product, and then I do a trading account, and then I, I have a commercial account, or whatever, those are all journey paths, that I can look for patterns and say, should I upsell Scott, should I cross sell Scott, he's been a good patron, I want to move him from a silver, you know, bank customer to a platinum one. And that means that I should give him a mortgage and other kinds of things. Those same patterns, recommendations and algorithms are what we use inside of fraud, to show the same path. And it's also very similar data to say, well, Scott was doing great, and then suddenly, he got a wire transfer from Mexico. And then, you know, he immediately worked with his broker, and then he prepaid the broker, and then the broker, you know, got paid the exact same amount of money. Well, what happened, right? Well, Scott, now has turned into a negative, like you mentioned that customer review. So now Scott is laundering money. Well, originally, we thought Scott was a good customer for a long time. Well, the difference in the graph database is I can find those anomalies easily, as Daniel mentioned, right. And then the way the pattern or the engine is that's running in the background, it's looking for things that are similar, or different, or radically different or similar, but patterns to people that are connected to fraudulent or SARS. And it's constantly sort of reevaluating, and training and learning that the human doesn't have spent all that heavy lifting time trying to figure out what those patterns are. And that's really kind of the point of this graph. thing is that it's it's a bolt on to what you already have, and it's not ripping out other systems. So it's a long answer to the customer 360 journey, and this fraud journey are very similar. And it's very helpful to you know, to lay persons that are trying to track all this right.

Greg Irwin 29:37

Scott, can I to jump through and I want to cover a lot of people's questions here. You asked if there's another use case that we can demo quickly. Are you able to do a demo on customer 360? Something that you can show the group is that is that set up in a way that would be would be useful?

Scott 29:57

Yeah, here we go. So in this case, What I've seen now is a traditional dashboard, this could be something that a customer service rep is doing. What I can show now is again, I'm finding things in this case like low sentiment, in this case, brand Farmington appears to be a valuable customer. I don't know how or why it appears that he's got low sentiment, so he could be a churn potential. In this case, I see that I have a dispute on a credit card. Well, that's interesting. I also see that I see multiple lines of business. So he owns a 30 year fixed, he's got some personal items, he's also got some wealth management products, etc. And then he's got some commercials. So he's got a business. Well, one of the things that we see here is using the power of the graph to show what is the next best conversation? What should I recommend? How can I help turn Fran into a gold customer? And what I see right now is that well, he needs to perhaps be guided to get investment property, or maybe some higher end wealth management. Okay, great. I also see that he's got a red indicator on his screen. And so what I can do now is I can start to let's see if this thing is hung up on me here. Sorry, folks. But basically, what's happening now is that, here we go, oops, I get unstuck. So now what I do is I clicked on him, and I say, well, his co partner also has low sentiment, well, that's now leading up to he's got a stolen or perhaps a dispute on his credit card. He's got his co business partner, which I found from the graph to say, there's low sentiment, well, now I can actually go look at time, which is this ability to say everything was going along just fine. He was, you know, he was doing all these great things, well, then he lost his credit card, well, that's bad. Right, then what I can start to do is I can do again, these are plugins for how we're connecting to the graph, so we can make recommendations, etc. And this is sort of that aha moment. Well, what I'm doing is while I get him on the phone, it looks like I could save him. In his personal banking, I've used the power of the graph to recommend very detail and prescribed home loan that can save him money, I can do things like wealth management, right? I can say, well, you're debt to savings, this is a very detailed kind of a portfolio analysis. So what I've done now is in this sort of forward view of customer 360, I can make recommendations. Now let's say I was an analyst, how did I get the graph to make those recommendations, I can go look at segmentation. And say, show me all the people who have a low sentiment or are potentially churn candidates, what I can do now is I can very quickly start to see segmentation, I can click on at risk. And again, I'm using the power of graph to do very deep link connected data. What I can also then do is look at that influence and say, well, Fran, is very connected to other high net worth individuals. Well, that's super bad, right? Because if Fran turns Not only does he have a wealth management account, and a banking account, and those other things that we saw, but he's also connected to Kevin and Tamara, and these are sort of mega investor and very influential people, again, our Moroccan kt analogy, right. What I can do now is I can ask the graph, where are those other connections inside of the entire bank trends, not the only one? Well, I see that done and reason all these other folks that are there. Now, what I've done is I've used that power of the graph and time, by the way down here on the bottom, that now I can create an intervention, right before I turn that customer, I can go create an action, which is what we saw on that other screen is recommend a friend while you get them on the phone, or maybe it's an old fashioned phone call, right or something like that. But that's, that's kind of, again, a quick scenario for what we did there. So any, you know, that's a lot going on. So the sentiment, it looks like. What we can do then is inside of a graph, there are things called features that you typically see inside of a machine learning. So for instance, I can look at how many times has he come to the website? How many times has he come to service, I can also do a voice sentiment on a call if he calls him I can see how many kinds of activities that he's he's he's responding. And what we typically find is people that have a burst of online activity and then call a call center to cancel their credit card, we can start to see that those things are bad. I can also see in his timeline that he used to use his credit card all the time, all the sudden he stopped using it even though he got up, he got a new credit card, so maybe he's hesitant. Maybe he's looking for another credit card, right. So those are all attributes on the data model that we can run our risk algorithm inside of TigerGraph to create that sentiment score. Yeah.

Greg Irwin 35:01

Hey, Scott, great stuff. And you you asked how is the load? sentiment determined? I hope we got the answer to that here through how it was through the triggers, or thresholds for that. Let me ask everybody who's on the line here, drop your questions or comments. What's one thing I'd like to hear? What's one initiative you all have for this year, if there's one wind that you're looking to implement for your team, Linda, shared with us it was the periodic reviews. There's some focus here on AML. For sure, there's some focus on customer 360. But put in what's one initiative that you're working on for your team or your team is working on that we can make sure that we incorporate as part of the conversation.

Scott 35:50

There's a there's a note here on the HTML for the account when I can show that one real quick as well, while everybody's typing something in real quick. So one of the things that we can do is we can create alerts. Now, alerts could be for customer 360, they could be for anything, quite frankly, they are anomalies that that we want a human to look at. And over here on the right, this is what we're building is we're literally building an alert. And back in our AML example, it says if there is a wire transfer, from a city, a high risk country, or it could be you know, pick your own could be from a city or state, whatever that is. And using a TigerGraph algorithm, I could say if there is a known account name that is connected to a known fraud case, or something like that, I can use something called similarity. degree of centrality again, those algorithms that we talked about Moroccan kt with, right? So if there is someone who has a similar account name that's connected to fraud, and or I could do something like number of transactions or frequency, you know, there's three, anywhere between $3.15 $100 greater than three transactions a day, within five days, right? All of these alerts become triggers for a human. And they could again, they could be over here and my dashboard over here, you see these alerts. That's how the human gets told what those things are. And in this case, this alert says, Hey, there were four credit cards and five devices that were connected to known fraud. You should do something about it human. Yeah. Right. So anyway, so that's that. That's sort of a long answer to the one that was posted in for alert. UK and AML. Yeah. So we do this all over the world, basically.

Greg Irwin 37:44

Excellent. Thank you. Thank you so much, Scott. Let's, let's keep going around I want to invite in. Well, if you're available, Quinn, Quinnita. Let's see. Are you on the line with us?

Quinnita 38:00

Yes, I'm here.

Greg Irwin 38:01

Well, nice. Nice to meet you. I'm interested because you're coming at it from a different type of organization than HSBC. So I'm curious from Celsius, what's what some of your perspective and some of your initiatives that you're that you're that you're bringing to this, this meeting?

Quinnita 38:19

So I, for those that don't know, I work at. It is. Basically, I couldn't say this in front of regulators, but we're a crypto bank. Um, that's the simplest way for me to put that, um, we provide some of the same services as a bank. So it's just easier for me to say we're a bank. However, in front of a regulator, we have to use very different terms, of course, right. So in AML, and my title is head of AML for, and I have a traditional banking background. So it's very interesting that things that I see one of the biggest things that we deal with, because my company is a startup, and which most cryptocurrency firms will be startup is lack of data, and finding businesses in the industry in the AML industry that are able to work with us, just because one thing that we deal with is cost. Because we are a startup we're looking for, you know, is it scalable. And then as we grow Celsius has grown tremendously in like the last year I'd say we went from like onboarding. Under 100 customers a day to now we're onboarding maybe 7000 customers a day. So looking for businesses that not also the cost benefits us but are they scalable? Will they be able to work with us through the long haul? With all of the limited resources that we have,

Greg Irwin 39:59

Excellent. Quinnita. So, first of all, congratulations. It's an exciting, it's an exciting spot in sector, you raise a couple key points that I'm going to bring back here to Daniel and to Scott and Michael, one question is, what about lack of data? What if you don't have the data sources? You're not coming at it with just a wealth of data? You know, how do how have you seen institutions? You know, kind of jumpstart their knowledge base?

Daniel Eaton 40:31

Yeah, yeah. No, that's nice. It's a it's a great reoccurring question for lots of companies is how do you? How do you augment the data that you have? And, and sometimes it can, you can, you can get some amazing traction from the data that you do have using graph because it, it kind of merges all that information and all those connections between the data to add a ton of value to explore and understand that in the crypto space, I posted a link in the chat to a company called Merkle science, who today has a 2.5 terabyte graph, entire graph that I think their data emerged from a bunch of public chain kryptos. They, they took all that data, and they they brought it into TigerGraph. And now they can explore and understand where transactions come and go. And using TigerGraphs, deep link analytics capabilities, you can actually trace transactions when when people think that they're being really tricky. Running a running a transaction 30 or 40 different places before returning home in hopes that you you cover your tracks, because Oracle and relational databases, and even the leading no SQL databases outside of graph technologies, can't they can't handle that type of data manipulation, because of the number of joins necessary to understand that data, we can actually do that natively very efficiently. So Merkle, Merkle science might be an interesting, interesting partner to look at cognita. We can also talk talking in more detail too, if you want to engage directly.

Michael Krogh 42:27

If I can weigh in, Greg, to your question. With regards to lack of data, right, this is a real concern for a lot of our customers, especially early on in their journey. for machine learning, deep learning training type of practices, one of the thing that Dell brings to our customers, part of the solution that I helped develop was our discovery workshop process. And this is a very iterative, pragmatic approach to stepping through an organization's general operations, your mission statements, your use cases, help identify further use cases. And then as we step through each one of the identified use cases, part of what we do, what our specialists do on our side, is to identify where there might be a lack of data. So this is where Dell really steps in from more of a general practice to help our customers along that journey to be in that true data driven business. So I just want to throw that in there. Please feel free to reach out. That's where we can, we can certainly help with that conversation.

Greg Irwin 43:26

Wonderful. Let's, let's keep going around a couple other stories here in our final 15 minutes or 10 minutes or so. And let's do this. I know people have other tasks, you know, other meetings. Let's finish strong here for our final 15 minutes. I I want to bring Paakow into the conversation. Yu. Thank you for your questions. Mika, I'd love to bring in for a story or comment and the rest of the group. So Paakow let's let's go to you first. Thanks for your questions. Please give a little intro to group.

Paakow 44:10

Yes, thank you very much, Greg, and thank you for inviting me to this group. So I am the AML Compliance Officer and privacy officer. And we are we are a broker dealer and investment advisory firm. And just for full disclosure, I'm here on my personal notes, whether they're representative of the company. And yes, as we as we progress into more remote type of work, and also transitioning from more brick and mortar type of meeting our customers to going online. Some of the risks and challenges that I see onboarding customers to the online on my procedure for which I believe can connect express another bed with regards to her hair type of company. And with this challenge, the first of all having to understand and know the customer knowing that you're the person that you are reviewing as the actual person, as well as any data to to reveal concern surrounding the customer challenges that I am currently expressing, exploring, in regards to changes to this the profile of the customer after they've been on boarded, that could make them a higher risk customer. So a customer might not be happy when they were brought on board, but then they could become a pet later on, and how these ongoing monitoring and graph can or any other software can help us identify these type of profile changes. So those are the angles I'm looking at.

Greg Irwin 45:58

I got is this, what's the urgency? In other words that I guess, as we've mentioned earlier, there's always continuous improvement. But yes, but But then there are things you have to do right now. Because the regulator's telling you because you have, you know, a threshold that you've broken through, and you actually have, you know, a management imperative. So, my question is, what's the urgency around know your customer? So yeah, customer improvements.

Paakow 46:30

Yes, ongoing, ongoing monitoring of the customer as a as a regulatory requirement. For in my space. The lucky lucky part of my company is that we are an introducing broker dealer. So we introduce a customer, the client, to an insurance company, a mutual fund, and we establish an account over there where transactions happen. So when it comes to transaction monitoring, there's more reliance on the mutual fund, and the other financial institutions that the customers are introduced to, however, there's a responsibility for Client Profile changes, which I still have a database of those customers to monitor. So the regulator definitely looks at all aspects of monitoring, including profile changes. So yeah, okay.

Greg Irwin 47:22

If I gave you a magic wand, and you could just magically make one meaningful improvement into your process, what would you What would you give your your organization, what would you enable them to do?

Paakow 47:36

I hope that we get some sort of alerts that when customer profile changes happen, whereby the customer becomes a concern for further investigation, get a notification rather than doing a periodic review of our customer base. Proactive, proactive engagement as opposed to reactive.

Daniel Eaton 47:57

And that's an that's a, that's a really great graph use case where, where we can, we can both understand how things are changing in real time, and provide a an alert based on various various thresholds, sometimes we can have logic built into the analytics to look at what what those changes could be. And when you hit certain thresholds, you can receive an alert if there are specific changes in mind. Other times, we can also apply graph and machine learning algorithms, once we start to understand some of the profiles of of those personas that you're you're interested in. And we can generate a vectorized representation of that and apply similarity algorithms. So that immediately once you have a new signature, you'll actually receive an alert and say, are these people have have now met a threshold that should warrant further investigation? And I think that's, that's, that's something that will be a very good graph use case. Thank you, Paakow. Thank you, Daniel.

Greg Irwin 49:19

Thank you. Let's go over to you, Yu. you nice to speak with Yu. Hey, quick, quick intro and tell us Well, tell us a little bit about your your current focus.

Yu 49:31

Right, so I'm not doing anything related to AML. So my I'm working for an insurance company. We have an IM support and analytics support to a group we call it digital experience, which the group to develop and maintain the all the online content or features to drive customer digital engagement. So you know, our goal is try to let people register online do a lot of things online, rather than call the call center. You know, they can be self sufficient and they can be, you know, have better experience with us so that it can stay longer. So what I'm, I'm trying to get is try to get some idea. So I have been using Tableau for many years I have been, I thought tableau, you know, graph, or the, you know, visualization feature is great. But what I see here is even better, you know, our things, showing here in the demo is pretty impressive. So, I'm trying to understand what other scenario or use cases can be applied through this. So from what I see is, a lot of data has been built, so that he can go to the granular level of customer, right individual, you know, very, the most granular level. So I was wondering, what's the aggregated level we can go? So for example, a segment of a customer, you know, there are things there. I don't know what other case scenario, you can kind of go through different Galli aggregated level so that we can slice and dice like normal analytics work, and we haven't done yet.

Scott 51:03

So there's two ways to think about that is that when you think about a hierarchy of things, right, so you know, in a supply chain world, you have, you know, a car is number one, number two, then you have the subsequent, you know, electronics bill of materials, which may have 400 parts in it, and so on, and so forth. The point of a graph is to have that atomic level, where you can roll it up to assemblies, and then assemblies turn into, you know, elements, and then elements turn into cars, right, you can use that same mental model for customers, for clients for anything that you can think of, it could be a product or a service. And so the beauty of the graph is that as that hierarchy gets built in the knowledge graph, keep gets better and better, because now I have more data. But you can start off originally, one of the other questions was around sparse data. We do that all the time, because we don't have all the data, but over time, it gets more and more enriched. And that's really what we're trying to do is that slicing. And dicing is something that graph is extremely good at it's called segmentation. And then we can interrogate it to say, is this segment similar or different than a pattern I've seen before, then that is a huge differential over a Tableau, Tableau, you have to go tell it what you think you you may find a, you know, our blue sweaters used in Texas, more than green sweaters, and you have to go in and write the SQL code. But in a graph database, I asked, I say, what is the best color of sweater in Texas in the winter, right, and it comes back and you know, shows me the empirical data and that I can make a decision. That is a huge difference in the way a graph database works fundamentally. And the way you know that other data, but that gives back to your question about, we do segmentation, and those kinds of things, that is part of what we do on a daily basis. And here's yet, I know you're tired of Scott's demos. But basically, that's what you see here is we can see a dashboard here where I can start to lean in on more of those kinds of things. And what I can do now is I can do segmentation, right, I can actually go look at risk, based on those segmentations, I can click on those, I can see where my customers have come from and where they are going to. Right. So these are technically those those nodes that we saw in the graph. But now what I can see as if they go along this blue band, that's a good thing. And then down at the bottom, I can see where they've turned into last. So I can sort of track where people are going. Here's that segmentation that we mentioned, etc. So again, all very good things that graphs can do at an incredibly more powerful way. Because the difference with the Tableau and some of these other ones is they're designed to show people things that happened in the past. Whereas a graph is designed to be I don't know where you want to go or the question you want to ask, but I'm ready. And when we're ready for you user to ask me a question that I didn't know. And so I think that's-

Greg Irwin 54:17

Yeah there's a good clarification there in terms of the difference of what you have here, as the capabilities of a graph database, as mentioned, versus leveraging a standard enterprise data warehouse with a business intelligence tool on top. That could be anything.

Daniel Eaton 54:34

That's a very, that's a very, very great observation, Greg, the the two things that that are huge differentiators for for us is the underlying graph technology. That that's the that's the engine that can provide those real time drill downs in exploring the data applying in in graph D deep link analytics queries as well as machine learning and graph algorithms. And we can connect up with any type of visualization tool we've got connectors for, for Tableau and Looker and others. Power BI is another great one. graph is something that, at the engine level enables that that snappy real time response and and also it enables data scientists ask those questions that traditionally, we've heard from lots and lots of sophisticated customers, that they, they couldn't ask their previous system, a number of these questions that with with graph, they can now get the results in real time. So customer 360 is is a fantastic use case. We talked a little bit about that. And, one, two cool things to note with Dell, because Dell has been an amazing partner with us. Dell is deployed into production at the world's largest healthcare company with us to do a patient 360. That's a customer 360 application is one of their top applications built on top of the graph. And it enables a unified view of the customer across 300 plus data sources. And, and it's used today in production by over 26,000 center representatives who who rely on this, this user interface and tool to make sure that they see a unified view of the customer. And it also provides recommendations to help them make sure that customer satisfaction and those call center response times and issue. Resolution times are as small as possible. And also one other cool thing with Dell is dell.com is actually going to be using TigerGraph for this this use case also, I don't think we can get into too many of the details around that use case. But it's it's another really, really awesome use case.

Greg Irwin 57:07

Look, let's get one more we just have a couple minutes here. Micah. Ah, I love the ski picture. Let's see if we can get you in for an incremental thought. And then we'll we're going to call it a session. It do us a favor, give it give a quick intro

Micah 57:26

It's Micah. But you couldn't tell by going either way? No, I mean, there's no way you would know, maybe that's the incremental ad I have is there's no way you can know about anything, but you can take your best guess we are just a startup startup BD, and we're gonna have a big ml lift in Asia. But we don't even have our primary vendor lock down. So this is like basically we're still in AML 101. And, you know, this is more of a graduate level.

Greg Irwin 57:58

So you're into, like, what is Parkside?

Micah 58:02

So we're going to be a broker dealer where people can buy us stocks, and the people are located in Asia.

Greg Irwin 58:08

Awesome. Cool. Good luck to you. What, how about one more Rebecca? Rebecca, I think you're on the line with us. I see a 253, which I think is up in Seattle. Are you? Are you on the line with us?

Rebecca 58:25

Yes, I am. Thanks. So I'm the first vice president over project management and it and I really don't have much of a play in AML. That's within our risk management and compliance department. So I know that's always a big topic in in banking, but they I don't I don't have a whole lot to add to this discussion. But I do appreciate the conversation and the information that's being shared.

Greg Irwin 58:52

Excellent. Rebecca, thank you for joining. We're going to wrap here we're at the end. Daniel, Michael and Scott. Gentlemen, are closing, closing comment and one for me. Big thanks for everybody for joining, we will send around the list of names, not emails. And again, as I asked, use it as an opportunity to connect with peers across across the country. Daniel, any closing comments?

Daniel Eaton 59:19

Yeah. Thanks. Thanks, Greg. And thanks, everyone for your engagement and discussions on the call. I definitely learned learn something and I'm happy to happy to connect with anyone who wants to see why Gartner says 80% of analytics and AI innovations in over the next five years will be done using graph technologies. So if you're not using graph today, happy to help you understand how you can be prepared for the future.

Greg Irwin 59:48

Thanks, Dan. Michael?

Michael Krogh 59:50

Yeah, you bet. And again, thank you again, for everybody joining this was a fantastic discussion. We're here to help out and right that the some of the folks that are on the call with us are free resources to you to reach out and partner within your organization. Learn more about graph, learn more about how Dell is helping our customers out again, free, free free, much like the commercial, we can help you. I don't want the big name scarier where we have a lot of we're part of the pre sales organizations name to get. We're at your service. So thank you again, everybody, for your for your assistance and participation.

Greg Irwin 1:00:25

Excellent. And I'm going to ask Scott Scott, the man with the the hands on the keyboard. Yeah. Any any closing thoughts for the group here?

Scott 1:00:33

Yeah. So you can you know, after the session here, we'll send out some reference links. But education is your friend. You can't learn graph in a day. And graft is not a silver bullet. So you heard that from me, can't solve everything can't cure COVID. But it can do an awful lot. And as the flip side of that is it can show you where COVID been and where it may go. So there are some unique things about it. But the links that we'll send around afterwards is really part of your educational link, if you want to do that, and then as everybody mentioned, we're here to help.

Greg Irwin 1:01:06

Awesome. Alright, everybody, thank you all. I look forward to speaking with everybody on a future call and follow on. Thank you