I'm from Ava company based here in Berlin with a with a large R&D office in Novi Sad just like half an hour 40 minutes drive from Belgrade up north and I'm here today to give you a glimpse into what we do at Eva without actually trying to sell what we do but actually I really trying to give you an idea of how things work because whenever we describe what we do they are usually some kind of questions coming up and I'm really happy like to have this presentation which allows me to give you maybe a little bit of answers to that it connects a little bit with what we just saw before like we have to the notion and their vision to create the safest world or a safe world a secure world based on deep data and information so the usual metaphor that we use one of the methods that we use to explain what we do is that we are kind of a weather forecast or meteorologists but not for wind rain Sun
and temperature but rather for crime terrorism healthcare related risk like epidemics but also civil unrest and political risk and all those kind of things that could actually affect like one's life or one's property with a little bit of difference to what you just heard when we talk about facial recognition this is not our field so our system doesn't look at humans or individuals our system looks at behaviors in general it looks at locations and things that are happening so to say an urban space for example or in organizations but really without going to the point was saying like pinpointing oh that was Thomas and that was Martina but really saying okay there is something happening and giving people who need that information that information in time so we're addressing a couple of things with with our systems so we talked about as I said crime and violence but also public disorder fire vandalism high-risk events but also large-scale events you know like when there's a huge football game taking place that might have an effect
on the security of a city or the security of a neighborhood etc and we we work for um for police for example in London the Metropolitan Police and London is one of our customers but we also work in public transport so there are a couple of public transport operators using our systems and their day-to-day operation centers to figure out like how the situation is and when something happens how to actually deploy resources to make the public transport safer and more secure or large logistic companies who actually want to be in the know and something happens on the world that affects global supply networks and global supply chains to explain it a little bit in a short version what and that's really really simplified of course so our system looks to find incidents so these are usual things that happen and people report about it or we have sensors etc I come to that but from all of these various sources that we are looking at we try to figure out what is going on and
then we're processing this information for that we are looking at a lot of different data sources and one part of the presentation will be to give you like a quick overview over how that functions and how many things actually around us exist that allow us to see how the security of a place is affected or can be affected even surprising things that usually people wouldn't even think about which have an effect on security of our surrounding of you know lack of organisations so one question that I'm that I'm able to raise since since since a while ago there was this terror attack at the at the London Bridge and and there was you know like things going on in London you may remember so we have been able to to see that happening four minutes before the traditional procedures and methodologies that the along the police has been deploying before they have seen that we know that because we during that time had a couple of pilot projects going on with several boroughs in London
and also with with with people at the city hall that are responsible for for crime and and and from there on it was kind of clear that we would work together so ever since like we're working with London police and the city hall to support them in the specific very specific activities no need to talk about that but it's of course not related like to individuals or humans but really about the security and the safety is a fabric of the city and so the AVA risk score which is one if you want to say like one of the core products or one of the core pieces of it is looking at security and safety from from the time dimensional in a time space dimensional perspective if you want so we have the past security of the past the historic a situation of a given location like how it's a statistic security of a place that can be relevant for example if you want to park a car there and there's a there is a there
is a history in that area that the park might get you know vandalized or stolen or something like that and would be no good to know that there's such a such a such a situation but also the present situation you know like if there's right now an active shooter and Alexanderplatz there are organizations and people that really need to know they need to know right now and not you know read about it half them half a minute or even two hours later on the news and then of course like the predictive part of it like looking forward into the future figuring out like actually what could happen that is also a combination of the situation that we have now in the history so all of these data both the past the present and the future they belong somehow together because more or less like really like meteorologists we can see specific contexts in a given space and figure out like if we have such context before just to remain in the case of public transport
if we know for example that in a Thursday at six o'clock in the evening usually we have issues on platform eight and not on platform 4 that has happened you know like for the last five months then it's almost common sense to say well next Thursday at around the same time the risk for something happening on that platform is much higher than on other platforms so we can actually use this data to go into like predictive modeling what we do is context sensitive so the idea here is that we are capable of of individualizing the security and risk score not only to individual people so it makes a difference whether I'm a woman or men of you know like whether I look like this or like that or whether I have this religion or that religion that is again like the part that we try to keep a little bit outside like a four hour focus because there are a lot of things you know ethically related to that but of course it makes a
difference whether you're an old pijl or a brand new Bentley right so also this like makes a difference so we can integrate that and actually give contextualized security information so again like just quickly to round this part of the presentation up we are designed in their by their by the principles that GD P R has been established or that that have been established through GD P R so we've been working very very hard to ensure that the system as such as really brings no headaches like to our customers when it comes to privacy and data protection which we believe is isn't it's an imperative like in particularly like with these technologies if we're deploying them if we have using such technologies to actually make the world safer and make the world more secure cities and families and organizations then there should be no doubt about like the fact that as related like do those kind of things so with that I would like to introduce you a little bit like to the way how we
get the information and how are you bringing it together like into one common system so there are a couple of pools of data that we are accessing the one piece of data or the one pool of data of course is the internal data of our partners of our customers right like if for example a police has the history of 30 years of of monitoring and tracking and documenting incidents in the city then this data as such has already a big value to understand like what the current situation and the future situation is but then of course internal data for example from insurances or for mobile network operators or from you know like organizations that actually have data that could be relevant for for armed security but then of course like so sensors in IOT we heard about cameras that can be used in in various ways not just to detect faces and understand who that person is but also to detect behaviors but in public places for example so to guide you a little bit
through the opportunities that are there thanks to data and an advanced and advanced algorithms and advanced machine learning and and all of these kind of technologies one thing that is really interesting is that when we of course give our our partners our users the opportunity or to also manually add incidents you know something happens I know it I can put it in so by doing that more regularly the system learns of course you know like and if we have some specific patterns emerging all over through the time then we can of course again use this data to to help to inform the situation but then of course like also incident in risk prediction so what you see here is like a like a photo that I took like from the from the screen of one of our science of data scientists this relates to the to the data pool of Los Angeles police so we took the we took the crime data from around ten years on an hourly basis so everything that happened in
during the last ten years for every hour of a day for every day of the week and we trained algorithms against that so that our algorithms would be capable of actually predicting the actual security for the day after for the days to follow so this is how we more or less approximate like all our machines and our systems to understand from that amount of data like what could be the outcome and we are pretty precise already which means we are running with an error rate of a little bit more than 5% so for example if we predict for the upcoming day for a specific neighborhood five stabbings so let's have 15 stabbings and it will either be fifteen or fourteen or sixteen but this is how close we are like with based on the information that we use also what is interesting that comes from the existing data sets is deep contextualization the fact that we take information into account that is not necessarily related to secure it to you right away opens up to
see things that all of a sudden seem to have a correlation or even a causation so there is this very interesting research very recently published by the London School of Economics where they found a pretty stable and and reproducible link between pollution and crime now people are arguing and discussing what is actually the reason why it's like that why is air pollution and crime so heavily related and they're you know like tons of concepts right now going around it could be something like a big a broken window a factor you know maybe you heard of that that kind of concept but it's a little bit like aspirin for us you know like we don't need to understand every time why something works if it works right like with I don't know how aspirin works I just take it and the headaches our way and in this case when I know air pollution goes up pity crime goes up that's enough for me to know right and we can inform based on that so sometimes it's
not even necessary to explain things that detailee digital media monitoring is one thing that we do it's pretty obvious of course you know like people are posting on Twitter and on Facebook and on YouTube and all kind of media and they're putting out this information and of course you can see when something happens and after all like when something happens in a in a city and a day to day life usually there's always someone who's the who's the first eyewitness and the question is what happens with that information in the first situation or like when it happens and in the past people of course they ran like to the next cell phone booth and called the police or something now they have smartphones but they have options on their smartphones right Thanks so they can call the fire department or the police but they can also take a picture and post it on YouTube and then the Facebook and Twitter and Cohen unfortunately or maybe even fortunately depending like how if where the police is
using systems like ours as many people are actually using social media to to actually get relief of the trauma that they're seeing you know like a building trashes you know like and the first thing that they that comes to many people mind is to make a photo right and that is not because these people are bad it's just because this is what we are used to we wake up in the morning we take the phone we tweet we check ours feeds all of that stuff so in such a situation we are not trained to do something else right like it's very traumatizing and therefore we we find it kind of kind of kind of crazy but also kind of kind of understandable that nowadays of fire departments without actually actively track social media they're always lagging behind you know like people are posting pictures of burning buildings and flames high up in the sky but nobody is calling the fire department because everybody things like someone else already called them right like an making photo
and all of these kind of things so for organizations like the fire department all the police such things really become important and like in this case at the at the at the Munich Olympia centre where we had the presumable terror attack which was actually just one young man who who you know like behaved differently than we all expected but I wouldn't probably I consider this really like being a terror attack as the way that for a minute or an hour or two hours all of a sudden people were thinking like well like this lady wrote there's you know like a gunfire everywhere around Munich even at Davos which was actually not the case but all of a sudden social media also becomes like a like a like a relief like a vent like a relief tool for people like to somehow communicate and we learned a little bit about about these past events and how to actually detect fake news for example so one thing that even doesn't come like - on people's mind but
can you imagine you know like you're standing in front of an active shooter or like an active shooting taking place or you are building up and fire the last thing that comes to most people mind is to invent a hashtag so for us in some cases hashtags are actually a very well marker for us to know that most probably this is not this is not an eyewitness because the way how people are behaving in such situations is really different like they're using these type of technologies much differently than if you would simply do that to gain like so Cayne awareness so hashtags much closely related to someone trying to get awareness and get shoulder taps and likes but no hashtags more probably not someone who's actually thinking about how can I get more visibility on social media but this is like really a different thing that is happening there and this is also like what I wanted to speak about here like really trust scoring and fake news detection which became a really important part
of our work of our day to day work like to really figure out how can we ensure that we don't inform like our customers our partners about things that are actually not true or that actually not real and this is something that we try to deploy like that was just one example but if you want like we really have for each and every single channel that goes into Ava we also we also have a have a trust scoring so to say up and running which automatically checks out like whether a specific Twitter account for example is actually trustworthy that can go even that far that we check how old is the Twitter account that had just posted this and what are the connections of the Twitter account does the Twitter account have a picture you know like those kind of things that allow us to see whether it's a bot or you know like those kind of things something that we experimented a little bit but that will which I'm also telling you to give
you an idea over how broad the spectrum is like how you can use such technologies and data without even interfering like with personal data really I mean with personal data is to look at network cells like what what what what kind of behavior would we actually expect if for example at Alexanderplatz at 3 o'clock on a Saturday noon afternoon all of a sudden the consumption of the key network cells are going low and the network cells around it like from the mobile network operators have a much higher consumption it at least tells us something about like how people are moving in this area you know like and it's an unusual way of movement and that is just today we're not even talking about 5g and and the low latency and high precision which is so to say the thing that we are already preparing for with that we can actually see how people are moving through a city and potentially also see anomalies that allow us to understand okay there is something now happening that
is really an anomaly people are running away from something so what is there so actually deploying further technologies and things to see something else have we tried out and which works perfectly but not in Germany in Germany it's not allowed if you do that in Germany you go to jail there's a phone kindness you know like even if you could technically listen to police alone trying it actually gets you into jail so to say and in the u.s. this is not the case we can together open up a website there's a look at the list of all the radio frequencies of of us police and fire departments and emergency dispatchers and we can listen like two police chasing bad guys in miami-dade County you know like so there will be you know like all of that is freely available on the internet you can listen to it like Spotify so if you take a if you take like a like a like a speech recognition service you don't even have to invent it right we've
used some alexa web service and transcribe what has been discussed there and you figure out that they're even using structured language you know it's not like people talking with each other but these are professionals using codes and no code to means go there but without without blue lights on and our code 3 is you know like how to behave you know like what kind of things to expect and what kind of incident has happened there so we even get structured data even though it's verbally structured data but it's structured so we transcribe it and from that we can use the same NLP like natural language processing technologies that we use to pass Twitter content to figure out like where something happened and what happened so that way we can even take the complete US American map and map it out like with real-time data from from from emergency and police and and fire departments as simple as dead right so it's it's it seems like in the first moment like science fiction but this is
something that you with a little bit of of technical background and a little bit of of time can actually do yourself right it's not really complicated something that is also really interesting is like ambient audio detection also a source that we are looking into a much deeper which means you know like how does how does a train station or Airport sound like when everything is okay you know you record the gender mark here in Berlin for 50 weeks and you will know what the normal sounds atmosphere is on that place and then you can then you can run the real-time data against that algorithm that you have created and all of a sudden you figure out that oh you know there people are getting louder you know and screams come up and maybe even gunshots which brings me to the next one which is another technology so now all these technologies would develop ourselves I have to say we are as Ava more like a central nervous system where all of these data sources go
into and are being contextualized so in this case for example we're talking about a u.s. American company that invented an algorithm that works based on an audio trinkle a triangulation it has been tested in Barcelona so if someone in Barcelona picks out a gun and shuts shoots at someone then if they're in that specific area the lamp posts I cook with microphones and because because the sound waves travel at a specific speed and if you have more than just one of these lampposts you can actually figure out by how long it took for the sound to reach a specific lamppost where the shot has been fired so that way you can triangulate actually the spot where the shot has been fired and not only that from the waveform you can detect which kind of gun that was so it was maybe a Beretta or it was an ak-47 no it was a Glock so that way you can see from the data this was not a police gun that has been fired right like it
gives you information like really in real time it puts it on a map and allows you like to deploy resources much faster than ever before and then cameras of course like you just saw like the awesome presentation around like facial recognition and those kind of things but this kind of aggression and detection is something that is really really interesting like alone in Berlin before gear like the Berlin public transport there have attender out of things they are deploying four thousand more cameras in the in the subway system they already have like ten thousands of cameras I think but now there additionally cameras being deployed and if you think of London is probably the city with the highest rate of cameras like installed everywhere but these are technologies that work after the fact so they don't help you when you are getting stabbed or getting robbed it only helps the policeman to go there with a you know like with a with a document from the judge to check out like the camera that was around
there so maybe they can get evidence from it after the fact it's not really helping in the moment when it's happened so if you deploy computer vision algorithms that allow you I think there's a company also in the exhibition area that is actually doing such kind of algorithms when you deploy such algorithms you can actually form the live feed you can detect when something is happening people are fighting someone is running away from someone someone is pushing someone you like those kind of behave that actually point to a to an aggressive behavior to a normal behavior and again that gives you the time like to really much quicker deploy resources and get people to where they have to be or tell people to withdraw from that area because there's an anomaly so those kind of things are possible and they are only possible like that because if you think 4000 cameras which human should sit there and look at these camera feeds right like there's no chance for nobody to look at that plus in
addition to check like all the tweets that are coming and plus in addition to check out all the news media that has been posted and to check like all of the sensors and fire censuses this is unhuman it's impossible but these people that are actually currently working behind the scenes to ensure that all of our societies and our lives are safe and secure they are overwhelmed with this data so they actually need something that allows them to put all of this information together and understand it in a much better context and this is actually what what is happening within within Ava so really wanted to guide you a little bit like to some other things like for example Nia you repeat patterns I mean this is this is criminology from the 60s right so imagine tomorrow I go and break in your home right or a new home and I steal your computer and your TV and your home so I steal all of your computers and all of your your TV sets and what
happens next day you're extremely angry you had a call with the police with insurance and all of that and you know then the next step will happen you will go to media markt on the online and buy a new computer and a new TV set so smart guy I was breaking in into your house will know that and they will return right because I know you have a new TV and a new computer so it's it's almost like you know it's almost funny you know like there are there are there like these bad guys they're just coming to steal this stuff away from you then to throw it away because they know a week later you will have the brand new Apple MacBook stand in there and then they return so based on these kind of patterns police is already working right so it's nothing really new here but by adding additional information and things to it there can be a lot be done so what we do is like we provide these these operator
interfaces they're usually being in accessed by situation centers so you know like public transport or on police so the people responsible for that they look at that they try to understand what is going on in their cities or in the areas that they're responsible for and but you also say like like buildings or locations and even insurances and underwrite as I said like we want to know like particularly like with bigger cities you know like if if a Lloyd's ensures the Olympic Games you know like they actually want to know what is going on there then you have like the field staff yeah yeah like the field staff right like running around they have right now they're blind yeah they're being called like to a specific situation and they arrived and they know nothing right they don't know what this history was of that area they don't know what is happening right there exactly and with that we are actually informing them and giving them the opportunity and this is something that we actually started
with the original idea was to say we give a security situation center to everyone to every single individual and and we right now we do that by an SDK so if you know like any company says like we want to deploy something like that like a security and safety app for tourists or whatever they can come to us and we can we can provide all the all the libraries and all the modules that are necessary to actually deploy something like that a really interesting it becomes when you think about how this data can actually influence our day to day life so I'll be really fast with that but this is you know like I don't need to explain it like your car telling you you should not park here because it's unsafe because of potential risk of being stolen or even calculating the route from A to B you don't want to end up with a 911 in the favela on your way to the airport in Brazil same goes for financial markets right like
you those financial markets that take very very fast and a bomb going off in front of Foxconn has an influence and the stock prices of Apple or of Samsung but also tourism and travel right who wouldn't want to have like a pro Ted like button to say I want to narrow down like the hotel's based on the security of the neighborhoods where they are because Fifth Avenue yes we know it's Tiffany's and Hamas and all the beautiful things in New York but Fifth Avenue is also South Bronx right it's just the street that is very long so if you might happen end up in five and you wanted to have like the luxury hotel and actually calling for uber someone is pointing a gun to your head to get your smartphone right so with all this being said I'm very happy that you took the time to take a little bit like a look on what we do at Ava it's usually something that we don't do like to be that open but if you
have questions or if you wish to help like a demo or something just let me know and I'm happy to give it to you thank you