Webinar: See what others miss. Multi-domain visualization for actionable intelligence

Transcript

Raphaël Alassar: All right. Hello everyone, and welcome to another Cambridge Intelligence webinar. It’s great to have you all here with us today. So my name is Raphael, and I’m the head of partnerships here at Cambridge Intelligence. Joining me today as our main presenter is my colleague Jan our product manager.

Jan Girman: Hello, everyone.

Raphaël Alassar: We’ve got, we’ve got a fantastic session lined up. Today we’re diving into multi-domain visualization for actionable intelligence. We all know the struggle. We are absolutely swimming in data from different sources and trying to make sense of it all at once. It’s incredibly tough to do that, so today we are going to explore how we can bridge that gap.

To guide us through all of this, I’ll be handing things over to my colleague Jan in just a moment. But first, a quick piece of housekeeping on how today’s session will run. So Jan will run through the main presentation first, and then we have reserved plenty of time to dive into your questions, so please don’t hesitate to drop your thoughts or questions in the chat panel on the right side, I guess, of the screen.

At any point during the talk, I’ll be gathering them up, and we’ll get as many as we can during the live Q&A at the end of the session. With that, let’s go into it. Jan, I’ll hand the floor over to you.

Jan Girman: Thank you very much. I’m hoping you can hear me loud and clear. Nice. And you can hopefully see my screen.

So yeah, thank you everyone for joining. This is a, essentially a rerun of the presentation that I gave at ISS World Europe in Prague last month. I’m assuming most of you didn’t get a chance to catch that, so yeah. It’s great to have you here today. So the agenda for our webinar today, I’ll, I’ll just give a, the briefest of introductions to who we are.

I’m hoping most of you know that. If you don’t, let us know in the chat. Happy to have a sort of separate session, you know, with you, kind of t- talk you through who we are and what we do. But yeah, after that brief introduction, I’ll sort of, kind of present our hypothesis on the, the challenge that is posed by the kind of modern fused intelligence approach, specifically for visualization.

And then I’ll talk you through our sort of, kind of product development approach to this. You know, how our, our various SDKs can come together to kind of tackle this, this, this type of problem. And we’ll sort of look at each one kind of individually and then we’re going to go through a few live demonstrations of some multimodal user interfaces that we’ve kind of built together using our toolkits for various kind of intelligence use cases.

And we should, as Raphaël mentioned have plenty of time at the end for Q&A, so please do drop your questions into the chat Right, let’s get cracking. So like I said, I’ll keep this brief. I’m, I’m, I’m hoping most of you have heard of us, that’s why you’re, you’re here signed in today. But in case you haven’t, the, the sort of the, the strap line is the one thing and the only thing we do at Cambridge Intelligence is build software development kits that help our customers build user interfaces for visualizing, interacting with, and making decisions from complex connected data.

Now the way we do this, as I said, is we provide a number of different software development kits. We started about 16 years ago now first in the graph visualization or as you folks in Intel like to call it link analysis space. That was with our KeyLines JavaScript SDK. And then we kind of grew that portfolio with ReGraph, which is a kind of React version of essentially the same thing.

And then we’ve also sort of added to the portfolio over time. So we sort of over on the right-hand side about six years ago we added in KronoGraph which is kind of designed for temporal kind of timeline visualization at scale. And then finally last year we launched for general availability MapWeave, our geospatial SDK for connected data visualization.

Now naturally as these are software development kits they can kind of work in whatever space, you know, and market you’re in. However, given our history where we started very much in that kind of intel police fraud detection kind of space with KeyLines all those years ago they, you know, all of our products and, and many of our customers have kind of been geared up for this y- you know, for, for building applications within that sort of intelligence law enforcement y- you know, sanctions geo and all these sort of, you know, kind ‘connected spaces.

And that’s, you know, kind ‘built up both in terms of the feature set but also in terms of the credibility and, and the sort of, you know, operational deployments of our customers’ applications that include our toolkits over those years. And again, you know, if you have any more questions about, y- you know, who we are and what we do on, on the whole, please do drop them in the chat or, or let us know and we can have a follow-up conversation.

So what’s the challenge when it comes to kind ‘fused intelligence? So, yeah, and this is something that I sort ‘reflected on a fair bit after coming back from Prague back in June. You know, it’s, it’s, it’s been another fantastic year another fantastic session at ISS. If you haven’t been, I can highly recommend it.

It’s a phenomenal event for you know, all things intelligence, you know, vendors, end users you know, a myriad of companies providing fantastic data sets. And the theme, a, a little bit to my surprise, but, but, y- you know, ultimately kind ‘unsurprising, I guess i- is still very much, you know, data fusion.

All the vendors are talking it, you know, all the system integrators are talking it, and the users continue having challenges with it. Y- you know, if you kind ‘put a- put aside the hardware folks at that event, y- you know the software space, I reckon you could probably slice it down the middle. You know, on the one hand, you had the vendors and the sort ‘solution builders, you know, kind ‘building various types of applications, which ultimately take different types of intelligence and fuse them together for the end user to kind ‘you know, consume as fast as possible and kind ‘draw conclusions, you know, make decisions, build their cases, whatever it might be.

And then on the other hand, you’ve got data providers, right? People with, you know, kind ‘web-based platforms or sort of, you know deployable data sets or, you know, just purely APIs, you know, a myriad of different sources of intelligence data. And the theme continues to be, you know, how do you bring all this together, y- you know, and, and into a coherent single pane of glass type picture?

Now, we’ve been in the space, like I said, for about sixteen years, pushing the narrative of link analysis and using, you know, tools like ours to kind ‘help users interact with data which lends itself naturally to be visualized with or walked through as, as networks. But these days, there is a myriad of other sources and types of data that one wouldn’t naturally think necessarily of as a, as a network, but they’re certainly related to it or, or could be, and that’s gonna be the theme of the conversation here.

So the type of things I’m talking about is like, you know, geospatial movement data, you know, time-based kind ‘you know, temporal data sets, patterns of life, you know, video feeds, you know, imagery. Yeah, the list goes on. So what’s, and, and the challenge which I think this poses is, is, you know, that while the industry has standardized on things like link analysis for sort ‘navigating some of these, some of these data, I would say that it hasn’t kind ‘y- you know, sort ‘met the challenge of, of fusing all these different types of int- intelligence, not necessarily into a single visualization, but into a kind ‘a seamless experience that continues to allow that analyst, you know, that investigator, y- you know, that person building up a case to y- y- you know, sort ‘navigate through the connected nature of these various bits of data in, in, in the best way possible.

So our approach to this problem, to this challenge is, is a multi sort ‘modal visualization one. And this is the reason we’ve been growing our product portfolio the way we have been. It’s perhaps a very simplified way to look at the world, but I reckon if you consider a lot of the use cases that you’ll encounter within the intelligence domain you know, the type of data, then the visualization can kind ‘cut across these three domains.

You’ve got time, you’ve got space, and you’ve got connections. And you know, just going through a few examples, right? This might be, you know, law enforcement data, things like, you know, how are you visualizing modern lawful intercept, you know, from a, from a mob- you know, a modern, you know, mobile phone, which no longer just gives you, you know, calls and text message data, but gives you location data, gives you, you know, social media data, all kinds of things.

Maritime intelligence, you know. You might want to track, you know, hundreds, thousands of vessels around your coast, but at the same time you want to be able to dig in to the beneficial ownership, which is a network of a particular vessel, or see the connections of ownership between different vessels doing suspicious things.

Pattern of a life, pattern of life analysis is another lovely example. This is, you know, this is spatiotemporal, right? You know, thousands, hundreds of thousands of mobile devices moving through a city over time. What are the patterns? What are the outliers? Financial crime, you know. This is, you know… It’s not just about y- you know, who’s sending money to whom.

It’s that, you know, exact sequence of these events that, that takes place that really matters and that kind ‘is important to building up a case like that. So we’re going to take a look at each one of these kind ‘briefly individually, and then we’re going to kind ‘go into sort ‘the the combined interfaces after that.

Naturally, we’re going to start with graph visualization, which you’ll hopefully know us all for. So this is, like I said, our KeyLines JavaScript and ReGraph React S- SDKs and amongst other things, these allow you to, you know, build experiences y- you know, for example for, for you know, navigating open source intelligence data.

So here we, we built this demonstrator with data from, from our friends over at Shadow Dragon. They gave us access to their open source social media kind ‘API, and we just looked at two sources of data here, the company’s house data and LinkedIn data, and y- you know, built out this, this little graph kind ‘navigation experience you can maybe spot there, kind ‘centered around David Beckham.

And y- you know, looking at, you know, being able to expand from that one entity into sort ‘all the entities th- in which they are involved or, you know, companies they own, kind ‘where, where they’re director of and then kind ‘continually being e- able to expand from that. So building up an experience where you don’t have to splash the entire massive complicated graph at the user, but providing you the tools to build experience that let the user naturally navigate and investigate this network the way that it best suits them.

Use our automatic layouts, you know, to kind ‘create ideal visualizations, which that can be, you know, exported or snapshots that can be cr- created and then revisited time and again. Really important when you’re sort ‘building up a case that you, for example, need to take to court, that ability to kind ‘r- faithfully reproduce exactly that view which led you to, to the conclusions that, that, that you’ve sort ‘arrived at.

And then also things like, you know, animating through changes in the network. Like, for example, you know, many of the solutions in this space, you know, offer users the ability to kind ‘reconcile what an algorithm might detect as identical entities. You know, so maybe you’ve got multiple data sets that you’re fusing together and you’ve got David Beckham’s name spelled differently in every one of them, right?

You know, you know, many solutions out there have, you know, algorithms that detect these kind ‘things, suggest to the user whether they sort ‘agree and want to kind ‘you know, fuse these David Beckham’s all into one. You know, and our layouts, our animations, these kind ‘… You know, they allow this sort ‘thing to happen very seamlessly within the UI, provide all the interactions and hooks for you to sort ‘do that.

So that’s just a very, very brief snapshot of, of our graph visualization capabilities within this space, but we’ll look at a lot more in due course. So let’s then move on to the time domain. So this is our KronoGraph product, and when we developed KronoGraph we were still very much thinking about the graph thing.

You’ve got, you’ve got nodes, you’ve got edges. But what we wanted to be able to do was, you know, in, in a, in a standard link analysis graph, you might have e- you know, entities, which in this case are individuals sending emails to each other. So if you’ve got 10 emails between two people, you know, you’d have two nodes, one for each person, and you’d have either, if you’re doing it wrong, you know, 10 links like an onion kind ‘between them for every email, or if you do it right, you’ve got one link with a little glyph that says 10 on it.

But what you don’t have a sense of at all in that visualization is when did, when exactly did those emails happen? Maybe it’s really important that those emails are actually being sent at 3:00 AM on a Saturday. Maybe they’re being sent to multiple individuals from one particular person. Maybe there’s a sequence of, of emails that get sent, right, between various people in an organization which actually help you build a case.

That’s what KronoGraph is designed to do, and it’s designed to do that at scale. So it will scale to hundreds, thousands of nodes, of entities, these names of people you see along the left-hand side there, and it will scale comfortably to hundreds of thousands, up to a million events on the timeline. And again, we’ll see why this is relevant momentarily And then finally and most recently, we, we– as I said, we released MapWeave, our geospatial SDK.

Say, for example, you’ve got two net– seemingly distinct networks or, y-you know, under a couple of sanctioned individuals. You know, these individuals own companies, those companies have various boards of directors, and the companies also own various high-value assets. And maybe a couple of those assets are vessels which had a ship-to-ship transfer event.

And you, as the analyst looking into these organizations, want to investigate that event. Maybe you, you wanna see where it happened in the world, what happened before it, what happened after. So MapWeave allows you to take those vessel nodes, and they are very much nodes, as you’ll see in a second, still in the network.

But now it can pl-automatically plot those vessels on their observations and sort ‘in– automatically interpolated trajectories for where they’d been. And you can use our time bar, or you can use KronoGraph, as you’ll see, to kind ‘you know, sort ‘filter through time and watch those vessels move along their trajectories.

And then you can take it one step further. You can say, you know, let me go… Let’s say you started from this view, and you know one of these vessels, but you don’t know the other one. And now you’ve seen them engage in a ship-to-ship transfer event, and you want to look into that vessel. Well, it could be as simple as clicking on it and changing the layer to see that same ownership network that we saw on the graph, but now spatially, you know, spread out over the globe.

Including nodes which don’t have to have geolocations. Including nodes which, like these vessels, might be moving, so they’re not in one static location as kind ‘time goes. All of that’s possible with MapWeave because it’s built on this concept of different s-you know, spatial layers. There’s our network layer, which allows, you know, kind ‘the link analysis on maps.

There is the observations layer, which allows you to then take those nodes and actually kind ‘plot them over time in different places. And then it’s all kind ‘brought together with a sort ‘standard GeoJSON layer that lets you do nice things like you see here. You know, color different sort ‘you know, exclusive economic zones of different countries and their waters, y-you know, and detect when, you know, vessels intersect them and, and so on.

So that’s MapWeave in a nutshell. All right. Let’s switch to some live demonstrations and look at how these come together for, for different use cases.

We’re going to start by looking at a anti-mon- money laundering sort ‘demonstration. So here we are looking at j- the sort ‘chart and the kind ‘the temporal bit, the KronoGraph on the left-hand side.

So let’s start with the chart. So we’ve got a number of different accounts belonging to various individuals and these accounts are exchanging funds. You know, so we can see the sort ‘the arrows for the, for the movement of funds. And it’s important to note that, you know, these, you know… Well, first of all, some of these accounts belong to the same individuals, and, you know, as, as if you’re aware of our graph visualization capabilities, you’ll know we can do things like, you know, put these, you know, different nodes into what we call combos, and that can automatically create what we call combo links.

So we’ve just sort ‘linked Woody’s, you know, multiple accounts into a single node and sort ‘aggregated the links automatically there. So that’s, that’s the on the chart side. And then the other thing to note is that, like I said, you know, we’re not showing every single transaction from Dragon Casino to Woody’s checking account.

We’re just showing the aggregate of all the transactions, including the sum. Notice what happens on the timeline, on the KronoGraph side, when I hover over one of these. So one of these links actually corresponds to three different transactions that we can see over there on the sort ‘center top of the, of the timeline, and so this is where the timeline comes in.

We still see the same groups, you know, the combos, so, you know, the, you know, like for example, Woody’s two accounts here. So all those nodes are kind ‘represented as these timeline entities, and you’ll notice as I kind ‘hover over them, they sort ‘light up on the chart. But rather than seeing the transactions aggregated into one sort ‘transaction between them, we see every single individual transaction.

Now, why does that matter? Well, for one, it allows us to do things like this. So I can… Because we can store sort ‘date/time ranges on nodes and links in, in the graph visualization, that allows us to then link that to KronoGraph and specifically its range event. And as I change that range on the timeline, I can automatically sort ‘hide, show, or foreground, background, the, the transactions and the individuals that are involved in that on the chart side.

Now, why does that matter for a case like this? Well- If we turn on the flagged transactions, right, again, ignore the, the timeline just for a second and just look at the chart. Because we’re looking into this network, right, we already know they’re, they’re dodgy, right? So it’s a little bit unsurprising that, you know, most of the transactions here are kind ‘turning red.

So in that sense, it’s, it’s not terribly helpful. But what I really want to know is, I want to see that money flow, you know, the sequence of the movement of monies throughout this network and how is it actually kind ‘making its way through, and that’s what KronoGraph enables me to do. You sort ‘see this, the sequence we’ve got here.

So let, let me zoom in on these initial transactions. Okay. You can see they sort ‘start from these cash deposits, and they go to these various business accounts. And now we, as we scroll, we sort ‘start to reveal, you know, how this money is making its way through this network, and ultimately, where does it end up going, right?

Because that, that’s sort ‘critical to our case, right? Who’s the, who’s the sort ‘ultimate beneficiary of all this naughtiness, right? And it’s this person down here. So hopefully that gives you a sort ‘a sense of the value that each of these visualization brings to this particular case, and that without one or the other, you don’t see the full, the full picture.

So let’s look at something a little bit more complicated. So this is our pattern of life showcase. So here now, we’re swapping the chart out, and we’re replacing it with MapWeave. And rather than looking at a sort ‘a network data set, we’re looking at, in this particular case, one million observations of taxis moving around the streets of Porto.

So this is, this is using the the, the famous Porto taxis open source data set, which kind ‘tracked about five hundred taxis over the course of a year. We’re just looking at twenty-odd taxis over the course of two months, and that’s sort ‘equivalent to a, a million observations here. And you’ll notice I can interact with every single one of these.

When I hover over one, I get a tool tip telling me which taxi and the date, time. And this is all in the browser, right? There’s no, there’s no backend running here whatsoever. We can efficiently load in this level of data and you can interact with it. Great. So what? Right? So this is where the KronoGraph piece comes in on the left-hand side.

So like I said, we’ve got here on the KronoGraph the entities are the taxis to which these observations correspond. And just like I showed you the time filtering in, in the graph, we can do the same in, on the MapWeave side. So both the network layer and the observations layer in this case, you capture the date times on, on those observations and that allows you to then with the KronoGraph API do things like this.

You know, as I sort ‘zoom in on a particular day, I’m seamlessly kind ‘filtering that million observations down to just the ones that correspond to whatever time range I’m seeing on the timeline here. So let’s run with that a little bit. So say, say I’m an analyst, you know, kind ‘you know, coming at, at this problem, pattern of life problem.

So there’s, there’s a number of questions or a number of approaches that I might be sort ‘y- you know, asking when I’m coming to this. Perhaps the simplest is I already know who I’m interested in, I just wanna find them here as fast as possible, look at their movements and see whether I can spot if they’re kind ‘you know, co-locating with somebody, convoying with somebody, right?

So KronoGraph gives us these built-in controls that, that allow us to do that. I can focus as it’s called, that focus control on a specific taxi, and in that case, I hide away all the other data, and I turn those observations into trajectories. So these kind ‘automatically interpolated tracks through the observations for that taxi.

And MapWeave API does all of that for me. And then I can do nice things like zoom in on one particular journey, and you’ll notice I have this KronoGraph marker here, which remains in the middle of my timeline, and that marker MapWeave has a corresponding now API. And as long as you tell it when is now, it can automatically place your network nodes, which don’t have a fixed location, but they have the same ID as these observations.

It’ll automatically place them on that spot for you. And as you change now, well, the– in this case, the taxi kind ‘moves up and down its trajectory. Why does that matter? Well, there’s also something we call pin in, in KronoGraph. So I can pin multiple taxis here, and I can start to do nice things like I said, you know, sort ‘see, okay, are there any– ever times when these taxis are kind ‘hanging out, you know, sort ‘you know, traveling the same routes, et cetera?

Obviously, you know, these days you probably want an algorithm to do that. But if an algorithm detects ones that are potentially, you know, kind ‘convoying events, as the analyst, I can toggle those on and exactly play through that and sort ‘you know, get the real feel for those, for those events happening, convince myself that they are real, right?

So that’s, that’s perhaps the easiest scenario where I know who I’m looking for. But very often, I don’t know who I’m looking for. Instead, something interesting has happened at a particular time And I want to see who was around, right? So let’s say someone of interest flew into the airport here in the kind ‘the wee hours of August the 1st, right?

So we kind ‘filter our data down and, you know, we can then we can use our GeoJSON layer in MapWeave, coupled with the sort ‘various event handlers to create ourselves a little polygon drawing tool. And when, when I finish dragging that out, you know, we create a spatiotemporal filter on this observation data set.

So youll notice now I have this, what we call a range marker here on KronoGraph showing me this is the time you’re filtering on. Here we’ve got our purple rectangle showing me the space we’re filtering on. And notice how a lot of the other taxis have disappeared because now we’re only looking at the taxis which were observed within that space and that time.

Now, now I can of course zoom out, I can see more observations, a wider sort ‘range of time, and that allows me to then do things like, okay, well, so s- one of these taxis potentially picked up my individual of interest, and then I know that person sort ‘left the Porto train station at some time the following evening.

So we kind ‘you know, narrow in on that time, zoom in on our train station here, create another filter, and lo and behold, that takes us down to just two taxis. So again, as an analyst, now I’ve got something to kind ‘go on, right? All right. Let’s delete our filters and let’s talk about the most difficult case, and that’s the one where My time of interest is very long and my area of interest is very large.

It’s an entire city. And what the hell do I do? Well, this is again where KronoGraph comes to the rescue. By automatically aggregating up into a heat map, we lose the sense of any sort ‘events that are kind ‘occurring across things, right? So if we think back to the email example, right, when we aggregate up to a heat map, we no longer see who’s exchanging emails, but we do see who has activity, right?

Who has events on the timeline. And we also see the inverse. We see who doesn’t have activity. So we can clearly see Taxi 364 either had a very long holiday or, or a very long service over here, right? You know, some gaps here for 483, another gap down here. So that’s if our timeline is linear. What KronoGraph also allows us to do is what we call scale wrapping.

This is one line of code, right? And you just sort ‘flip the mode of, of KronoGraph, and you’re still looking at exactly the same observations. But now, on the timeline, you’ve wrapped those observations around the days of the week as they kind ‘recur throughout your data set, and that gives you a different sense of pattern of life.

You know, it shows you that actually Taxi 304 doesn’t normally work during the sort ‘daytime hours and evenings on Sundays. However, if I zoom right in, you’re going to notice that there was this one event, this one time that they did a journey. Now, that to me as an analyst is an outlier. Right? Why? Why do they break their pattern of life that one time to go and have that journey?

So now I’ve got them pinned. I’m gonna switch off scale wrapping, and I’m going to start to investigate. Well, where else did they go, you know? Maybe I can figure out where’s home. You know, maybe I can figure out if they’re regular meeting with somebody else, et cetera, et cetera. So hopefully that gives you a sense of how, you know, together with KronoGraph and MapWeave, you can do large scale pattern of life investigations.

And th- there’s very little custom UI tooling and things you actually need to build to make a really powerful application. A lot of what we do here just uses the existing built-in events and controls. All right, let’s take it up one more step in terms of complexity.

So this is a kind ‘a law enforcement example. And here we have, you know, kind ‘combined our graph, currently very simple, just one node, our KronoGraph and our, you know, geospatial visualizations all in one. So let’s talk through this. So we’ve got here a device that belongs to our prime suspect. Now, they are a suspect because they were observed, you know, kind ‘in the vicinity of an incident of some kind.

And what we’re looking at here is we’re looking at their sort ‘final approach to, to that incident based on the movement of their device Now you’ll notice that when I zoom in, we’ve got this little range ring around this device, you know, something we kind ‘created with the GeoJSON layer. You know, sort ‘I don’t know, can’t remember what I set the radius to be, let’s say sort ’25 meters or something like that.

And so… And we’re going to be kind ‘dragging our timeline or, you know, we can drag the phone as well, doesn’t matter, the kind ‘MapWeave trackers work either way. And we’re gonna drag them along, along this approach, and we’re going to be looking to see if they interacted with anybody that might also be of interest to our investigation.

And the way we’re gonna do that is every time an observation comes into our range, and that observation belongs to a journey which completely coincides with our suspect’s final journey, we’re going to light it up. We’re gonna light up that observation here, and we’re also going to sort ‘add that connect- temporary connection in on the graph.

So in this case, we’ve sort ‘detected device 473 was in, in this area around this time. When I click on them, I will, again, rather than seeing the individual observations, I will draw a trajectory automatically through them and actually plot where they were w- down that now line, that marker. So when our suspect was here, actually turns out this particular person or this device was all the way up here.

So probably not of interest to us. Let’s unselect them and keep going. And in this way, we’re gonna keep going, keep going, and you’ll notice we’ve got this loitering zone. So we’re gonna kind ‘ignore every potential interaction before that, ’cause then we’ve kind ‘observed that our suspect spends quite a lot of time in this loitering zone.

So the hypothesis is, you know, they’re either waiting to meet somebody and kind ‘travel up together, or maybe they’re picking something up from someone. Now, here’s the other thing, fun thing about MapWeave. We don’t do the base maps. We’re not the base map guys. That’s not the value we add. MapWeave gives you adapters to your existing map, base map providers, and one of those could be Google Maps.

Now, why does that matter? Well, it means that you get access to all the sort ‘existing features from that, like in this case, Street View. So if I’m interested in seeing, well, what is this area that this person’s kind ‘loitering in, you know, I’m able to do that here. Great. Okay Quite a busy street.

We’ve got a station over here. There are lots of people coming and going. We can clearly see that on our chart, right? Like, devices kind ‘keep popping in and out. But we’re not gonna sort ‘select any of them. We’re going to wait to see if they actually repeatedly hang out with somebody. So I, I wrote a little, little sort ‘you know, function within my UI, which just sort ‘every time one of these happens, I kind ‘log it, and I’m waiting to see if somebody repeatedly kind ‘hangs around my, my suspect.

And as we sort ‘get to this kind ‘quieter area and then sort ‘head out towards the incident zone, we should see… Ah, there we go. Somebody was indeed detected by my little script. Device four one nine repeatedly detected within range of the suspect. So this is one of our annotations, right?

These can be kind ‘automatically drawn via the API, or you can kind ‘give the user control to sort ‘draw annotations around your nodes and as well as events on the timeline. But I’m gonna acknowledge that. Done. I’m gonna toggle on device four one nine. Ah. And lo and behold, right, they are well within the range of our suspect.

And not only that, if we scroll back our timeline to where their latest journey started, we’ll notice it very much started in the loitering zone. So it looks like our suspect waited there, met up this, this additional device, and then they kind ‘traveled all the way up here, and look where their journey finishes, in the incident zone.

So we have a new suspect, someone else that we can now investigate. One of the ways we can do that is we can go and look at the previous journey that they were sort ‘… You know, we’ve got them highlighted now, so that’s easy. We just go scroll back. You know, where were they before they met our suspect here?

Looks like they were sort ‘over here. So that’s interesting. They were sort ‘approaching the loitering zone and then, I don’t know, dis- disabled their GPS or, you know, maybe that’s a, just a glitch in the data. Who knows? Well, let’s follow it to the start to see where they kind ‘set off from because that could be interesting to us.

Let’s go up here. All right. So up in this area they started the journey. Let’s go and see what’s over there A tram station by the looks of it, and looks like some kind ‘hotel or something up, up over there. So yeah, maybe that’s something that I can kind ‘now follow up in my, in my investigation as I keep going.

So that’s it for that demonstration. Hopefully that gives you a sense of how these three visualizations can come together for a complex sort ‘you know, kind ‘spatial, temporal, kind ‘connected sort ‘challenge. So let’s go back to our slideshow. So we looked at these various examples. So let’s, let’s wrap up and then head over to any questions that you might have.

So as we’ve discussed, you know, fused intelligence continues to be a key theme for intelligence solution providers as well as for the end users. It does in our sort ‘… You know, from our perspective and from what we hear from our customers and partners, you know, provide a challenge in terms of visualization, you know, that goes beyond just link analysis.

You know, how do you bring these things together? And I think there’s a, there’s potential future opportunity here for kind ‘greater standardization on some of these, some of this user experience yeah, and with that, thank you very much and over to any questions that we might have.

Thank you so much, Jan. That was an incredible run through. To kick things off, I’d love to take us back to a real world setting. So thinking back to ISS in Prague, where we were exhibiting recently and where you delivered your presentation, how did this concept of multi-domain visualization actually originate with the people you spoke to on the floor? So were they facing the exact kind ‘data fusing challenges we talked about today, or did something else stick out to them?

Yes. Raph, before I get to, to your question, I think we have one in the chat from Anis. So can the system be deployed independently on the end user platform, or does it require a continuous connection to a central server?

Very good question, Anis. So our SDKs are purely front end SDKs, so they’re quite agnostic of the type of application that you want to host them on. I mean, obviously they are browser-based, so that’s the one kind ‘key requirement. But it does mean that they don’t care whether you deploy them as a SaaS application, whether you’re deploying them, you know, kind ‘in a hybrid model or whether, as in the case of many of our government and intelligence customers, you want to deploy it in a, you know, in a complete siloed air gap, you know, no connection to the outside internet.

It works in all of those cases. How you choose to serve up the data to those visualization SDKs is really up to you. The only requirement is that it’s, it is the format of, of JSON that, that we ex-expect, right? And that’s all very clearly documented. It’s a very simple JSON structure. I don’t know if that kind ‘helps answer your, your question. No, Anis. Perfect. Yeah, great. You’re welcome. Yeah, you’re very welcome.

Sorry, Raph, back to your question. So you know, how did, how did this message resonate with people at ISS? I mean, look, really well i- and, but I would, I would say that where I think it resonated the most, and actually most of the people kind ‘attending and, you know, asking me questions at the talk were whom I would consider end users, right?

So you know, again, I-ISS is a fantastic event for kind ‘interacting with a-various end users from various domains in intelligence and, you know, they, they absolutely get this, you know, and, and they have this problem and they’re seeing this challenge. You know, they, they’re faced with a myriad of, you know, different signals intelligence and, and, and other sources.

And, you know, they do genuinely have a problem to kind ‘draw conclusions from, from all the data that’s being given to them. It resonates with the vendors too. But I think as, as, as, as, as we were planning to kind ‘discuss here, like, you know, a lot of the vendors are currently very preoccupied with, with AI, right?

I mean, that leads me perfectly to my second question. So because as we all know, you can go to any industry event right now without AI being the absolute center of gravity. So how much did AI come up into those conversations and audience questions in Prague? And I’m really curious, what’s the general feeling out there regarding chatbots and AI assistance in our industry and, and specifically when it comes to visualization?

Yes. And again, I think this is… There was, at least in, from my perspective and, and keen to hear if, if, if you, if you heard sort ‘otherwise from, from, from partners. But I mu- from my perspective, you know, there was again a big difference in terms of the way the end users are kind ‘looking at AI and, and the sort ‘you know, how they might want to leverage it, versus, I think the way the vendors, the system integrators are kind ‘looking at this.

I mean, in terms of the vendors, you know, you know, you’re spot on for mentioning, you know, chatbots. Like everybody and their auntie, you know, is, is integrating a chatbot into their into their solution. You know, I think it’s the kind ‘the … I don’t wanna be unfair, but there’s a little bit of the, the unimaginative, you know, thing to kind ‘add in. What’s the quickest, simplest thing we could add? Let’s just pop a chatbot in, right?

Where I, where, where, you know, some of the things I’ve seen and, and the people I’ve talked to that I thought were really kind ‘comparing, compelling on that side were, for example, right, especially the platforms and the solutions that have been around, you know, as long as us or, or, or, or longer, they’re complicated beasts, right? There’s a ton of functionality, rich feature sets, not to mention the myriad of different sort ‘signals and intelligence that are kind ‘being brought in for the user.

So put yourself in the shoes of a completely new, you know, user of this solution, right? Like these things can be overwhelming, right? There’s so much going on and I’ve seen some fantastic ways that vendors are using, you know, a simple chatbot or a kind ‘a very similar interface to just help that newbie user kind ‘navigate these solutions, get the most out of them for their particular questions, their particular cases, like as soon as possible.

And where I think, you know, where we had some very interesting conversations, and I think the way that this market kind ‘might go, you know, you know, in terms of how, how, what the role that visualization plays in. You know, right now most of these chatbots, the responses they’re giving are, you know, primarily verbal, right? I mean, that’s what LLMs are kind ‘designed for. You know, people are using them to summarize lots of data, lots of reports. They’re using them to generate reports, you know-

You know, thinking more in terms of that kind ‘government, NATO, et cetera, space, there’s a ton of, you know, standard formats for reports, you, you know, and et cetera, that, you know, are an absolute pain to kind ‘produce every time. You know, these bots are fantastic at kind ‘just taking the stuff you give it, you know, and turning it into that. You know, s- but, but where I think there’s an int- interesting and potentially very valuable, you know, again, coming back to that kind ‘decision support, you know, sort ‘use case is y- the point at which they don’t just spit, spit out some words, right?

But maybe they draw the visualization for you, right? Maybe they annotate the existing picture you’ve got there on the map, on the link analysis, whatever, with the annotations, with the things, draw your attention in, y- you know, based on the sort ‘the, the, the, the, the learning and, and, and processing that they’re kind ‘doing in the background, and I think that’s a particularly exciting space.

I think on the end user side, there’s rightly a lot more trepidation, right? There, you know, there’s … You know, you, you’re talking about people who You know, a lot of what the decisions they make, they might have to stand up in court at some point in the future and justify those, right? And stand by those. So there is a strong sense of ownership of the ins- conclusions drawn and, and the decisions made, right? Whether that’s, you know, a, a commander making battlefield decisions, whether that’s y- y- you know you know, a financial crime investigator kind ‘y- you know, laying out their case, you know, for, for, for a particular investigation. Like, y- you know, these, these are serious decisions. They have to feel that ownership.

And so I would say that from their perspective, and from what we’ve kind ‘… what I’ve heard, visualization’s not going anywhere. It’s one of the tools that actually helps them validate the stuff that these solutions previously using, you know, more traditional algorithms and, and statistical methods, and now kind ‘more and more using AI, you know, is to validate those. Go and check it for themselves, you know. Sort ‘sift through the noise and kind ‘you know, focus on those few nodes, those few edges, those few observations on a map, and kind ‘reassure themselves that these are the correct conclusions or that they’re not, and then they need to go and do the manual legwork themselves, right?

But y- you know, I would love to hear from, from, from other people on, on, on the call today if they sort ‘… if this kind ‘perspective resonates or, or if you’ve s- if, if you feel different. I mean, this is just, these are my impressions from, you know, kind ‘that event and then sort ‘talking to, to, to users, people.

Ju- just adding to this and aligning AI to our toolkits. I mean, from… I, I work very closely with our partners and they love our MCP service. Okay. So when I introduced it to them a few months ago, they got super excited. Right … why don’t you take us through our MCP service maybe for those people on, on the webinar that haven’t seen them before?

Oh, yes. Oh, and a very nice comment there from Maria. Thank you. Visualization helps you tell, tell a story to the users. It’s definitely key for less technical users. I 100% agree. Yes, so all of the toolkits you’ve seen today they all come with their respective MCP servers. These are sort ‘hosted by us in the web. If you are a customer of ours, you’ve got access to them.

For those of you who not, aren’t in the know, the, these servers are a way to tell your code assistant. So this is something not targeted at your end users, this is targeted at your developers, right? So if, if your developers are using our toolkits to build out frontends for your applications, they can po- you know, point their code assistant, whether that’s in VS Code or in Claude Code or wherever they might be kind ‘y- you know, using an AI assistant to, to build solutions.

They can sort ‘point their, their working environments to our MCPs, and it just helps the AI kind ‘understand our APIs. And, you know, again, that specific format of JSON that we require for our visualizations, so they can look at that automatically and they can just sort ‘you know, churn out code, which will sort ‘work natively with our visualizations.

It’s a, it’s a massive enabler. We did one of these public webinars a few months back where a, you know, a colleague of ours, Kevin live built one of these triple visualizations, like on the call from scratch, you know? I mean, you know, we have great APIs, we have great documentation. You know, it’s, it didn’t take a huge amount of time to do that manually, but you’re still talking the order of days, you know. Like, whereas he did that in one call in thirty minutes, you know, and, and, you know, they kind ‘interacted across each other, you know, showed data across all of them. You can, you can find that sort ‘you know, link to our webinar and the associated blocks on our website. I think it’s it’ll tell you everything you need to know about the MCPs

Thank you, Jan. I guess, I guess someone is typing in a question, so we might wait a minute

Coming in from various people. We’ll just give it a second. I, yeah, I, I did wanna sort ‘throw that question, you know, you know, back at you a little bit, Raf. Like, you know, be- besides the MCP servers, like you know, how is this kind ‘you know, this multi-domain, multimodal kind ‘visualization r- resonating with our partners? And maybe you could say a little bit about wh- what it, what are our partners, who are our partners, what can they do?

Yeah, so I mean our partner ecosystem is, is, is, is very unique. So the vast majority of our partners are system integrators, solution providers, tech consultancies that are building mission critical platforms, bespoke end-to-end systems and solutions, mainly to the public sector, but also to the private industry. So that’s spanning across law enforcement, defense national security, intelligence agencies financial crime, and critical infrastructure projects all over the world

So when they talk to us about why they rely on the multi-modal view, they emphasize two main realities. So the first one is the, the single angle view is obsolete. So in today’s threat landscape, relying on a single type of visualization, a single angle, or a minimal amount of data sources is simply no longer valid. So modern agencies are completely overwhelmed by the sheer volume of data that they ingest daily. So if an analyst has to jump between isolated software screens to look at phone logs then switch to another app for mapping and another for timelines, they lose the thread. So connections get missed and critical response times are delayed.

So that takes us to the second point, which is the power of the three toolkits in one view. So this is why our partners rely on a fused multi-modal approach. So they think they need link analysis, timeline visualization, and geospatial living dynamically in a single view. So it allows an analyst and investigator, so those are the end users of our partners, to seam- to seamlessly move from one dimensional data to another without losing the context.

So with our graph visualizations, it shows them who’s connected to whom. So that can be criminal network or infrastructure links. The geospatial view immediately maps out where those entities are operating in a physical world, and the timeline lets them filter the exact window of time and events to see when and how those physical movements and connections interacted. So by giving agencies or whatever the platform is, the ability to pivot between the who, where, and the when within a single interface, so our partners they are helping them turn an overwhelming mountain of data into immediate and actionable intelligence. And we’ve seen that over the past few months and year, how’s it moving from A single view or a single an- angle to a multi-model point of view.

Nice. Nice. Well, I’m glad it’s resonating. Yeah, and we did, we did a fantastic webinar with one of our partners Future Space recently as well which for those interested, I’m sure there’s a recording kicking about somewhere we can kind ‘share with you. But yeah, that was that was really great. Completely unprompted by us, they, they, they, they demonstrated on the call the, the, the sort ‘multi-modal visualizations that they build for a kind ‘a, a, I think it was like a drug smuggling type s- type scenario. And, and I think there was a sort ‘a cybersecurity angle as well. So yeah no, really good stuff.

Well, I think unless there are any more questions… Ooh, we got something from Brian. Just curious, are you able to protect these truly innovative concepts so that CI and its partners can make the claim it’s the only set of toolkits that can do this? Well, yeah, there’s a fun topic of protecting JavaScript there, Brian. I think it’s you know, there’s a reason we got a sizable legal team. I, you know, I think I’m not sure we’ve, we’ve made that claim but you know, from what I’ve seen out there in, you know, in terms of Graphviz capabilities at least, I think y- you know, we, we’ve got y- you know, the sort ‘the, the fullest suite, you know, of the, of these kind ‘complimentary visualizations.

Obviously there is an entire open source realm y- you know, where all of these things, apart from I reckon KronoGraph are sort ‘available in some shape or form. You know, then it’s a sort ‘a standard, you know, build or buy type, type question that, you know, if, if purchasing an SDK isn’t justified, well then yeah, I mean, it’s not a, it, there’s not much to discuss at that point.

I want to add on how powerful our toolkits are how it can help you scale what you’re working on, how customizable they are as well, and one other point that makes us special that the people who are developing those toolkits are also the same people who are supporting our partners and, and our clients. So that saves the developers of whether it’s our clients or partners a big chunk of time waiting or searching for, for, for where can they find an answer for this problem or, or this challenge. We’re very reactive as well to, to, to, to all of our clients and partners.

Nice. Thanks, Raf

I see, I guess we can take one more question. I see someone’s typing.

Well, let’s see. I’ll give it another minute, but yeah. Otherwise, thinking about helping developers tell the right unbiased story. Do you provide any lessons learned in the docs from your experience, perhaps akin to GDS patterns?

Really nice question, Tim. So there’s a couple of ways that we go about this. So our when, when you, when you evaluate our toolkits or when you become a paying customer you get access to our, the SDK sites for all the toolkits that you’re licensing, and those include a myriad of examples of how you can use the various features. They also include, you know, these showcases that I’ve been sort ‘demonstrating now. They’re all kind ‘our production showcases, so they kind ‘th- they show the power of these things, and they’re all very much based off of the things we hear from our markets, you know, from the conversations, the type of datasets, you know, that our kind ‘customers encounter.

But then you also get access to not, well, you know, our commercial development team, so they kind ‘support our customers throughout the pre-sales process. And you know, they’ve seen it all, right? Like, in terms of the, the, the types of challenges, types of data. So they, they’ll … And they are developers themselves.

And then our support model, you know, for licensed customers is our, our developers. They are our support team. There’s a rota for every single SDK. Th- there’s a person, you know, always on call. Y- you know, and they are the developers who build our stuff, who’ve again encountered, you know, bugs, issues, feature requests from, you know, a myriad of our customers over the years. So they bring a wealth of experience to kind ‘support you with this, with this development. Great question, Tim. Thanks.

All right. Well, I guess that brings us to the end of our presentation today. A huge thank you to you, Jan, for sharing such great insights and taking us through those live demos. And of course, a massive thank you to all of you in our audience for taking the time to join us today and asking questions. So if you would like to explore how these model visualization tools kits can fit into your own systems or workflows, we would love to connect, so you can reach us directly through our website.

We’ll also be sending out the link of the recording of today’s session over the next couple of days, as well as we’re gonna be posting it on our website. So feel free to share it with your teams. So thank you again for your time today. Have a fantastic rest of the day, and we look forward to seeing you at our next webinar. Thank you. Thank you.

Thanks, Raf. Bye everyone.