Jim Webber with Dr. Edgar Osuna: Using Graphs to Take Down Fraudsters




Last year’s NODES event concluded with the enigmatic Jim Webber – Neo4j’s Chief Architect and CTO – interviewing a series of graph devotees, users, customers, and community members. Interviewees included a risk modeling scientist, a civic tech practitioner, and a software engineer.

His final keynote of the day was with Dr. Edgar Osuna, the Chief Data and Analytics Officer at TODO1 who received his PhD from MIT. (Woah. Smart guy.) Together, Jim and Edgar explored how graph technology is used by TODO1 to identify bad actors in real-time financial crimes, and applauded the Neo4j community for supporting and elevating their work. Check out the conversation between the two doctors below. Enjoy!

Meet Dr. Edgar Osuna, PhD from MIT


Jim Webber: I’m absolutely thrilled to introduce my final guest of the NODES 2021 conference. My friend, Dr. Edgar Osuna, PhD from MIT, right Edgar?

Dr. Edgar Osuna: Yeah.

Jim Webber: Actually, I feel a little bit anxious about this. I’m quite used to being the doctor. In fact, I used to have a Neo4j hoodie that said “The Doctor” on the back, but I think MIT kind of beats Newcastle. So now do you get to be the doctor? Is that how this works? I’m not sure. It’s intimidating.

Dr. Edgar Osuna: To be honest, I kind of normally downplay that. I don’t have that in my LinkedIn line. I don’t play with titles that much. I think whatever you’ve learned in the past and you bring to the table is what really matters – not the actual titles – but thanks for that intro.

Jim Webber: So let me dig at that a little, Dr. Osuna, while we’re at it. You’ve been doing operational research for a long time, which has led to where you are now as the Chief Data and Analytics Officer at TODO1, but the first kind of LaTex style academic paper of yours that I read was on support vector machines. Do you remember what they are?

Dr. Edgar Osuna: I do, I do. Yeah, it was very fun for me to work on that in MIT when that technique was still crawling out of Bell Labs when Bell Labs represented what it did in the past in terms of technology where getting first drafts out of Vladimir Vapnik, who was just an absolute genius in furthering that technology. And I was thrilled to just be in the right spot at the right time and bring a small grain of sand into what then later was built as a mainstream technique.

Jim Webber: That’s marvelous, humble as ever. So look, since those days when you were a young student full of optimism and support vector machines, things have moved on. You are now Chief Data and Analytics officer at TODO1. Now, I understand Chief Data and Analytics Officer, but what’s TODO1?

TODO1: What They Do and How They Use Neo4j


Dr. Edgar Osuna: So TODO1 is a company that’s been out there for over 20 years. Although people see us as technology – we don’t see ourselves as a technology company, we see ourselves as business partners of our banks – we service banks. And so we use technology to get them to derive business results, particularly I would say in two main areas: digital channels – so older web and apps of this world are run through us and are operated by us. And that, as you can imagine, has been a tremendous responsibility in this COVID era.

But also in the security space, because it’s coupled with that, right? You want to have customers that trust their channels and trust your applications so they know they’re protected. And of course, there are bad actors that try to take advantage of the distance that they now have with an app. In the old ages, they would have to come into a branch holding a gun. Now, they can do this from everywhere with a lot of more comfort. So we bring those two things into play and we do it to help the banks achieve business results.

Jim Webber: Okay, impressive. I mean, it certainly seems a lot more peaceful than robbing a bank at gunpoint. But I also imagine the kind of scale of say, for example, fraudulent activity could be so much higher when you’re doing it digitally rather than having to physically walk into a bank and hold up tellers at gunpoint. So how does that kind of evolve in your world?

Dr. Edgar Osuna: Well, it evolves pretty rapidly, because when you were doing it at gunpoint and you missed or you failed, you were risking a lot of consequences. Now, the fraudsters only have to get it right once. If they fail, they’ll just continue, and that’s unfortunate. It’s something that is part of the business here. And they evolve.

And just as we have companies Chief Data Officers and people working in analytics, I’m sure they have their own people doing the research and investing in technology to do whatever they do with a much higher level of sophistication. And it’s particularly, I’d say, more dangerous, because now payments are running into more of a real-time flavor. So payments, we want them to be fast. In the industry, the customers are demanding that to be faster. And obviously once the money leaves, it’s harder to stop. It’s not as before, where you would have more batch processing types of things. So a lot of things are changing and those are bringing challenges to us trying to protect.

Jim Webber: That’s interesting. It’s fascinating to think somewhere there’s an evil Edgar. There’s your counterpart who’s trying to figure out how to defraud your system, and you’re literally in an arms race with evil Edgar at this point. So tell us about how you came to graphs and the graph stuff that you’re working on at TODO1?

Graphs for Fraud Detection and Prevention


Dr. Edgar Osuna: Well, we were definitely trying to revamp our fraud detection and prevention system. And the first thing that popped up was, okay, so this is not a traditional database type of problem when you’re looking at representing knowledge and representing and calculating features particularly in real time. And if you think about it, when you go to a whiteboard and you start drawing a fraud pattern or fraudulent activity, you do that with nodes and arrows, which basically is a graph. So why would you try to do that with just tables if it’s not the real paradigm for us to talk about, lest to calculate stuff? So having been faithful users of traditional databases forever, how do we break that barrier and get into a different paradigm with graphs, particularly for a real-time system?

Jim Webber: Now, that’s interesting, because earlier in the conversation with Julie Fisher, she said very much the same thing, right? For her, I was looking at spreadsheets and I kept being able to see through the data – I saw through the matrix as it were, and she could see the connections in the spreadsheets, and that led her to graphs. So it seems that you’ve had a similar kind of a path, where you’ve looked at patterns and figured out that that’s hard to capture in relational databases, for example.

Tell me more about the real-time aspects, because when I’m here in the UK and I wanna send a payment, our faster payment system is great. It will send money from one account to a neighboring account in just a matter of hours, but I think you were thinking about something a little bit lower latency than that, right?

Dr. Edgar Osuna: Yeah. And the payment systems in a lot of the banks that we service in Latin America are real-time payment systems. So once the transaction is clicked the money is gone. Actually, I remember seeing that with a taxi driver who was actually explaining to me the payment system that he was using. He was definitely a believer in the QR payment system, and he was thrilled to say, “Hey, look at this, you do this and you see my account, here’s the money.”

So it was instantaneous in that perspective. Now, obviously that’s a challenge, because from a user experience point of view, I don’t even have the luxury of seconds to analyze your transaction. I have to do that in a sub-second time, so that all throughout the customer journey isn’t affected. And although every transaction goes through our engine, people don’t perceive that and don’t get the delays that you would get if you have the luxury of spending five, six seconds to do that. Throughout the entire journey it has to take less than half a second.

Jim Webber: So hang on, let me try and put this in perspective in my own mind here. You’re saying that kind of at continental scale, I can do instantaneous payments that to a human really do feel like they’re immediate? And as part of that workflow, you are doing some, I guess, some rather sophisticated fraud checks?

Real-Time Transactions in 21st Century Banking


Dr. Edgar Osuna: Yes, and in those sophisticated checks that’s where you guys come in. Let’s say, all of the memory, all of the understanding, all of the patterns, if you will, in history of behavior, in the map, and all the biometrics related to my end bank customers – all of that is stored and continues to be stored real-time just as we speak, and it’s used to derive whether that bank is going to challenge you with a, I don’t know, a one-time password, or let you go and get that transaction flowing. Or denied a transaction, because the confidence level is so high that we definitely do think that this is not Jim making the payment.

So it has a profound impact on the customer satisfaction and, therefore, all those digital channels for those banks. And in this era that is critical, particularly because if you think about it, you’re one Twitter away from being harassed to hell for a bad service or bad timing, or taking too long, or denying too much, or all of that. So it becomes a critical component for the banks in terms of their user interaction.

Jim Webber: Wow, that kind of brings it home, I think it makes it rather scarily real that even adding a second or two to a transaction suddenly becomes poor customer service.

Dr. Edgar Osuna: Unacceptable. It would time out. Definitely two seconds is a timeout. They’re automatic. And so, yeah, it has to happen and it has to happen at a scale, because if you think about it, two in the morning is not the same as, I don’t know, noon or early afternoon. When people get their payrolls, obviously there’s a lot more payments. When you are on certain holidays or Christmas, transactions go off the roof.

One of the keys to our investment in Neo4j was to have something that was not just flexible in terms of being able to manage all of that, but also resilient, so that if something goes wrong, because we are operating in a public cloud and things do go wrong every now and then, that happens, the system – the show – has to continue and there’s no excuse for us to get to the banks or the banks to their customers. And when we’ve had those issues, we’ve seen that Neo4j has performed perfectly well. Everything that we had, that we tested, that we thought from a theoretical point of view would work, things have worked and the system has recovered later without impact on customer service.

Jim Webber: Well, that’s very good to hear. It’s my ambition to make Neo4j boring. I think graphs are very exciting, but I think operating machinery should be as boring as possible.

So, look, a lot of what we’ve discussed throughout this session has been very much from a community point of view. I realized that you’re slightly different, insofar as you are a commercial user of Neo4j. How has it been as a commercial user, using the Neo4j commercial side, I guess, interleaved with the stuff that you work on with the community and getting community help, and all that stuff. How does that work for you?

How Community Support Makes the Whole Thing Work


Dr. Edgar Osuna: Well, if you think about how everything started I would say pretty well, from the way we got engaged from a business perspective, right? The sales process was a process where customer satisfaction was first, second, and third, and only when we were satisfied we started talking about, okay, so, do we need to sign a contract? I think we need to pay for something. So the main concern from Dave [Richard, Sales, SE] and Alex [Rivilis, Sales Engineer, SE] was “What is it that you need? Let’s understand what you need. Let’s keep all this working.” And then we talked about engaging that.

I would say afterwards, you obviously need the product to perform, which has been the case as I mentioned before. And I think you also need the aftermath to be handled very well. On the technical side, these projects, they’re never over it seems. They’re always evolving. And that’s where Rodrigo [Haces, Customer Service Architect]] and Susan [DiFranco, Customer Success Manager]] are bringing their expertise as well and keep on tracking what we’re doing, what are the resources that we need, and so on and so forth.

And if I may mention something else that I think has been critical for us has been the participation from the folks from CMS [Cloud Managed Services], because the reality is that Neo4j is a new technology, at least for us it was and continues to be. And operating that, I sleep very sound at night, because I know that the folks from CMS are operating and managing our cluster and all of our environments, including QA development. And I strongly recommend – if you really are operating something that is time-sensitive like it is in our use case – I’d say having the peace of mind of having people from CMS manage it, it works really well for everybody. So I think it’s been a, I know it sounds corny, but it is a team effort from the Alexs and Daves of the world to the folks in CMS, to help us have success with our customers.

Jim Webber: That’s a really gratifying thing to hear, Edgar. I think one of the things that makes me so happy about Neo4j is we have this wonderful community, many of whom are on this session even this late in the day. But the people inside Neo4j, we’re part of that community too, and it’s full of really nice people. You mentioned Dave Richards and Alex Rivilis. I know those folks very well over in the U.S., and they are so community minded, right? It just so happens they are on the Neo4j payroll, but they are so minded to help. And then Susan and Rodrigo, similarly they are there for you all the time, because they are members of the broader Neo4j community and I love that.

Dr. Edgar Osuna: They’re members of my team. When the customer feels that they’re members of your team, I think that’s when success happens. And we try to do that, because we have customers of our own, right? We feel that business partnerships that are built on trust and mutual trust are the ones that generate relationships that are lasting, right? Not just signing and fulfilling a quota. And I don’t think that cuts it.

Jim Webber: That’s wonderful to hear. I know that a lot of folks at this conference haven’t necessarily dealt with the commercial side of Neo4j. So I think it’s really nice to hear from a real customer that the commercial side of Neo4j operates very similarly to the community side. It’s full of people who really want to do well. That’s pleasing. And I do hope that Dave, Alex, Susan, Rodrigo, and James [Freeman, CMS lead] are tuning in, because they’re gonna be absolutely pleased as punch that you’ve called them out.

Let me zoom out in the final moments of our conversation. Edgar, we’ve known each other a while and something you said before really stuck with me. And sort of quite counterintuitively, that emerging economies – like the global south have an advantage, a competitive advantage, over those players in the global north. And that sounds very counterintuitive, but you’ve got a very compelling explanation for that around technology and levels of capital and so on. Would you mind expanding on that? I think it’s very interesting.

Industry Insights from an MIT PhD


Dr. Edgar Osuna: Yeah, I give you two pointers into that, because I think it’s a very deep discussion. And if you think about the software industry or some pieces of the technology space, they’re not capital intensive. So there are areas where not so developed economies can actively participate and the barriers are not as evident again for capital intensive types of activities. And you see some of these countries where education has been a big investment and the top talent that I have, for instance, for example, based out of Columbia, it’s at par – if not superior – to other talents I’ve seen in other areas in the world. And there’s also a very high level of commitment to the result. They’re not looking to move to the highest bidder after six months of working at a company.

So you have certain conditions that are good for developing certain parts of industry. Costs for a lot of those resources are also significantly lower. So there are competitive advantages there to take advantage of. By the way, I think Silicon Valley is starting to take note of that, because we’ve seen some of these markets heating up, but it happens that way.

And the other thing that is characteristic of some of these spaces is that for instance, if we talk about the payment industry, which we were talking about just a while ago, customers are used to something that is real time. They’re not gonna take batch payment NAFTA style, or I’m sorry, NACHA style, or fed style, where there are three or four windows during the day – like why? We’re already used to real-time payment systems. So that keeps those resources thinking in an advanced setting, and then when you’re reaching markets like the U.S. market, for instance, there I would say in some areas, we’re well ahead of the curve.

Jim Webber: That makes so much sense. I guess in a way, you’re able to completely skip the mainframe era, the relational era, and you’re able to pick modern technology tools that can really propel a business forward, I think, because you folks have.

Dr. Edgar Osuna: Yeah, and I think in our case, it’s been key. And I joined TODO1 only two years ago, but as I said before, this company has been going for over 20. All C-levels have been doing banking for over 20 years. And that means, again, it’s not a technology company, in the sense that we’re not selling bits and bytes. We understand how a bank works on the inside – not only on the technology front, but also on the business front. And I think those use cases are the ones that can add value to our customers.

Jim Webber: That’s wonderful. Look, Edgar, I have to say, thank you so much for taking time out of your day-to-day. I know that you’re due to go on vacation soon, so I really appreciate you especially taking time out to spend with us. It’s been wonderful hosting you. Thank you so much.

Dr. Edgar Osuna: A pleasure talking with you, Jim, and hope to see you physically at some point, hopefully before the end of the year.

Jim Webber: I hope so. I still owe you that cold beer that I promised many moons ago. I will make good on it. It’s a pity I can’t do faster beer payments. Thank you very much, Edgar.

Dr. Edgar Osuna: Thank you for having me.

Don’t miss out. We have lots of events coming up. Register today!

Check Out Our Upcoming Events