 Okay, good morning everybody, so I'm says game. I'm from the from Intel in the city of office I'm doing system architecture and today. I'm gonna be Presenting to you guys the work that Sunco and Cinderella They have been doing within the team. They couldn't come so I I'm coming in and representing there in the world that they do within the city of office and And any x so today we're gonna be talking about about health care about Mac about the edge edge to cloud security What are the foundational elements that we are working on in terms of edge architecture and and basically we're gonna I'm gonna be looking at some architectural elements that you are looking from hardware and and software perspective and The directions that we are looking as a company in this in this area So Let's let's get started a little bit on the on the problem statement or the or the context of the of the presentation today So if you guys I don't know how much you are into health care If it's your day-to-day job or or your day-to-day is modern orchestration system architecture So regardless to that it's I'm just gonna briefly touch on on some of the interesting opportunities that are happening in in this in this area So if you are familiar when you go to the to a hospital, right? So you're going to the different processes from admission to to the treatment to when when you go home continue the The process may be at home. There's labs analysis Doctors lot of papers lots of analysis and if you look at at how today's hospitals are Transforming there are many of the things that maybe two or three years ago We're done by by someone looking at at some analysis or maybe processing some data today All those things are being Automated right or on the process of being automated and that basically goes from the as I was saying like We'll talk about that later, but and natural language processing Or do to x-ray and and AI analysis for for a machine, right? So there are multiple areas in a in a in a healthcare domain where you could apply Technology to make things more effective and more efficient And just as an example so two days ago. I was talking to a to a doctor that he's Basically doing imaging all all the all day long and he was telling telling me that in the in the in the hospitals today They give they're giving them an amount of minutes to process one x-ray and depending on the complexity of that patient you the time Various right and and sometimes is is complex for them to keep the throughput that they have to do and obviously During the during the day they get tired and and and they can't get everything So any thing any method that allows them to do this type of processing faster or Identify things that at the first chance they wouldn't look at That's all that's that's very critical for for for for them So them obviously like when you look at healthcare and in general many of the H Deployments and technologies. There are some key elements that you need to always consider in terms of the architecture and in healthcare specifically anything that relates into data privacy and and Data anonymization. It's one of the key elements for for health care, right? So if you look at the at And the technologies that will be talking through today So there's a lot of emphasis on the on on how do we treat data? How do we keep data secure and how do we? Secure those things with the hardware and software altogether, right? And obviously like the other aspect that happens specifically with with healthcare is that For the nature of the data you can't get all all the data go into the cloud for example and do The AI learning on on the data from some of the hospitals, right? So think like us as federated learning where you need to distribute the learning of the models in the different hospital premises and then Combine the different models to create one one single model are elements that are for a foundational from the healthcare domain so When you look at the at the technologies that that you that we are looking or let's say like the workload or foundational microservices or services that that are usually used in this type of use cases You go from things like natural nlp type of algorithms that you do text-to-speech NLP for processing like questions or stuff like that or even analyzing the the voices and the documents You you go to computer vision obviously like for x-ray analysis even when you do like surgery there are some Algorithms that are being trained to identify when there are some objects that go in in wrong places and obviously you have anything that is preventing of For forecasting potential Even adverse events that may happen right and and one of the interesting aspects of of the healthcare is that there's large amount of data That if you process it in a you know in the right way and you create the right models They can help Substantially in optimizing and preventing some some things to happen So I think for me one of the interesting aspects in in this Specific domain is that what is the data monetization? It's going to be very very important For the healthcare domain in general because there's a lot of data that is data, but it's not you used for for for things that can basically improve like the Overlies but also like the logistics on the from the healthcare and Hospitals and and healthcare facilities in in general So When you look now connecting a little bit healthcare them, so I went through a little bit on on different areas from the from the healthcare domain where like Compute and AI and This type of technologies can help we went through a little bit some of the pillars that you may see in terms of the Warlords or or foundational microservices that you could use now when you look at the H computing and marring this with with the healthcare Why it's can help right so I think that one important aspect that I just talked about it on the so we talked about data Privacy we looked at we talked about Federated learning so you need to have this Continuum of of H2 cloud because obviously if you do everything on premises the cost and the OPEX Of the deployment is is high So you need to have some balance on on your architecture and see what do you do really on premise and what you can do on the Clouds right and for doing to that you have to have orchestration capabilities that look at the at the problem from end-to-end perspective the other aspect is Networking is obviously important You need to protect the data from the device to where the service that is processing that particular data How the bytes are moving from one place to the to the other and the other important aspect is in the end on the Modularity and how you treat each of these individual elements that create the the whole the whole architecture one important aspect here to to think about is that When when you look at all these use cases in principle, they could look like a some Monolithic piece of software But you will see and I'm gonna talk about that in in few slides is that more and more the The use cases or the warlords are getting the compost in micro services And that allows you that that basically some micro some parts of the world that are Microservices that maybe are processing or preprocessing some sensitive data You can do that on on premise and maybe the out The output of that data or that micro services that is not it has no Let's call it like critical or or private data. It can go into the cloud for example and and do the second level of processing and also like you need to have as well when you decompose the These use cases in micro services and then map them into to the proper part of the of the of the tears of the edge Now you have to have as well orchestration that provides the right level of Reliability that is needed in in this type of use cases right and an example is that you can't basically have one patient that it's being monitored or for You know in the ICU or stuff like that obviously like the the monitoring has to be continuous And you have to have failover mechanisms in case like some parts of the of the of the architecture They they stop working or connectivity does does not work and and here It's it's important that you need to at the end of the day everything is is a trade-off of of cost and It's a trade-off of of of the of of the security and and privacy that I was just talking about right then and one of the Important as I think in general when I see how the ecosystem is moving as I was saying like cloud native is becoming more more relevant for for for us for the work that we do and as I was saying this there's more appetite to leverage like standards and Leverages as Ecosystem the solutions that we that we work on and that's why we're like putting our bed in in in cloud native And if you see I'm gonna talk about the smart edge, but we are using cncf Kubernetes distributions and we want to basically Embrace all these Standards that are done by the community and contribute with with our Enhancements in in Kubernetes plugins Kubernetes operators and and and and and things like that one one important area of of of edge and especially in the healthcare is that it's not only about the the the compute boxes like a Dell server or a Super micro server that runs on premises, but there is one aspect that sometimes we Forget about it. We jar the devices one of the challenges that we we see in in healthcare on In in hospitals in general is that one of the complex complexities is how you connect all these different types of devices that are part of the of the hospitals and And it's not my day-to-day job But when when I talk for example, I was in in Bangalore one one month ago and one of the challenges that we That we saw talking with the with one of the hospitals that you are partnering with that They have a lot of brown brown fill Devices in the hospital. So how you integrate all those devices and and process all the data that they have is one of the Big challenges that that we have today, right? And the second obviously is anything that go related into data privacy and as as I was just talking about is how how you really make sure Make sure that the data is processed when it needs to be and that the that That all all that is on is data privacy is secure, right? The the other aspect is it's on the on the connectivity So we have the devices we have the data privacy and the other aspect is on the on the connectivity, right? So we are looking when From the architectural perspective how we converge different types of of accesses 5g 4g and PLS Wi-Fi and all these different elements that are part of the of these deployments and and see basically how how you connect Like the devices with the services and and how you make that in a in with a common common architecture so When you look at I don't know guys if you are familiar with a smart that job and so that's something that we We are doing from from Intel perspective. It's a it's a it's a Kubernetes Based distribution where basically we're trying to provide a Framework that you can use to connect as I was saying a few seconds ago like different devices with different types of of connectivity Technologies and then running that on obviously Intel platforms and over and the hardware that we have and Try to provide some level of what we call reference implementations that these are cloud native Container based implementations that we have containerized and divided in micro services so that you can get it and basically Or you get the whole thing or you just get bits and pieces that you need for for your application one of the things that we are heavily working on it's it's it's try to see how we are More and more standard with with with the ecosystem obviously when you go into Kubernetes Like Kubernetes APIs are standardized and well understood But for example when you go into the more Application part of the of the site and you want to provide micro services How do you how do you? expose and and provide the Microservices to the to the community that's something. It's not yet clear. I mean if you look at the definitions of microservices out there are Lastly coupled and stuff like that very very generic So we we're trying to see how we can contribute from the from the Application perspective as well and that's what you see like what we call the the Intel a smart edge portfolio so When you look at the at the smart edge The the way that you can think about it is as I was saying you have on the bottom the the cloud native foundation but then what we're trying to do it's on on this Baseland platform, we are trying to add new building blocks that basically allow you guys to get Reduce the time to market for using some Some types of technologies that are needed for the for edge architectures, right? And here obviously you have like elements of Security, so I'm gonna cover that in a second But you know that we're working on on SG x tdx technologies that basically Protect the the the microservices that are running on on a platform from the hypervisor and in other other types of attacks you have all the tpm, which basically is the Doing at the station on the platform so that you when you boot bootstrap the system You know that the system has not been compromised and obviously you have all all the all the different elements that we have in our platforms and technology that go from from telemetry we have lots of Different data that you can collect from the platform there from the performance monitoring using you need that you can use to be a standard to like Prometheus Grafana and and other types of of collect the another types of telemetry Building blocks and obviously like anything that goes with with with acceleration, right? So there is in in in the context of healthcare very net in general, so element media AI are building blocks that are heavily utilized so we are investing a lot of time on on Defining the right level of of hardware technologies that go from Isis that are exposed by the CPUs that goes with GPUs that have Systolic arrays for inferencing plus media building blocks so that you reduce the the TCO when you have media plus Plus a type of of workloads or even if you have to do certain types of analytics And you have to use for example a bx 512 that all those things are provided us as building blocks and basically When it goes into Kubernetes or providing the right level of of Kubernetes plugins Kubernetes operators so that you don't really need to handle with the complexity of Deploying for example an Intel GPU right so that's something that for for you should be more or less Transparent so that you get the hardware you get the the operators you get the plugins and it should be ready to go and it should be multi-tenant right There on top of these basic building blocks that that you can get you can utilize and you can include in your Kubernetes Distribution we are defining what are the experience kids. Okay, and then basically the experience kids are You can see this as a reference implement implementations that they couple together some of these building blocks so that you can deploy them the full architecture configured for those experience kids Examples are the private wireless kid. So what we do I don't know how much you guys have been working or dealing with integrating different flex run or Flex on plus core just like UPF type of architecture But in general is something that requires a lot of heavy lifting. So what we do from Intel perspective is provide this Private wireless kid that it's it's something that you get you can deploy in your Kubernetes distribution And you will have a set of run slice flex run Technologies plus UPF that you can get on and deploy into into your your system Something important is well and when on top of that you have the edge services that these are the microservices that I was talking about so for example, there is some Some something that is called the the L streamer pipeline server. There is a microservices a microservice that we provide there is Is a micro of service that you can deploy in your in your H node and you what you do You have restful APIs that you can register the pipeline that you want to process that may have I don't know Detection plus classification and then you using the same restful API you you can register different streams of cameras, right? And that basically and you will get the the result of the of that pipeline for that stream And you don't have to basically know what hardware is underneath and how that's managed This is done by that micro service, right? And you have any other edge services like that Network functions stuff from media and any so that you can basically Leverage all the stuff underneath in a in a more transparent way so Well, obviously as I was saying so the this is just a snippet on on what could go within within the smart edge developer experience key that includes Elements that go into the into the Kubernetes master obviously then on the H node and we provide different Optimized CNI's we provide different and optimized Kubernetes plug-ins Kubernetes operators and this is something that you can get and and install into your system from from from a scratch basically we provide all the ansible script that do that allow you to have To boot and configure a system in questions of of few few few minutes You can this is so this is an open source so that it's on the github so you can access to it You can download it and and you can test it out in into your systems It's you will see that there are some of the experience kits which require some more Agreeable contracts, let's say with with a smart edge open But everything is something that you can you can download as you will see So there are lots of building blocks here that they are open source components We don't have to reinvent the wheel So as I said so the the the aim is to be as much as Align with with the ecosystem so you will see that basically there are a lot of things that are open source But the work that we do is basically being sure that we provide the right level of of New elements so that we can expose the the benefit of our platforms and into the in into this Open source a building block so for example in terms of Prometheus So we have the we provide the right level off of Prometheus exporters for the hardware that we have the telemetry So that this is something that can use for orchestration so The the other aspect is okay now going into the into the implementation of the applications So I don't know how much you guys are familiar with open vino But basically open vino is an AI framework that you can utilize for doing developing and training your your building your Enants and you can optimize them on top of Intel hardware And then you can deploy them in your in your in your system architecture, right? And that's something that you can do Using the dear streamer pipeline server that I was mentioning or you can use all other microservices such as open Open video open vino model server that this is the the one the microservice that you can get basically and put your Your models in this in this model server and you get everything Optimized within the within the same within the same container one Important aspect again is when it comes into the deployment So I'm gonna say again container container container So we're really looking into into this microservice Kind of foundation because that's that's allowed allows you to take benefit of all their Kubernetes constructs for auto scaling For service meshes and stuff like that, which is something that you need for when when you need to deploy flexible And more scalable Architectures and in this case the up at the the model server is something that you can use for Implementing your own applications and that we are heavily utilizing in this in this healthcare domain as a Building block that you can use for doing the the the media and an AI inferencing something important that's this when you when you use this type of microservices You don't necessarily need to know what hardware is underneath So that's something that the microservice take takes care of and the beauty of that is that now you have the same The same application that can go from a from a nook that has core to another platform that has zion and zion sp using AI Isa optimized or you can go into another platform that has Zion D plus GPU right and that's going to be abstracted to your application so In the content now Connecting all the things together right so what we have done and and we will provide a little bit more of Reference at the at the end. So now you connect all these different pieces on the Intel hardware a smart a job and and the different Services or microservices that are meant to optimize and reduce the time to market for the application Development plus the connectivity part of the of the architecture and now you can connect all the different Elements to make the end to an architecture right and in this case that goes As I said that they being it can go from the patient sample scanner Doing that into training models that gets deployed into the edge servers and that implements the specific use cases that are needed for the for the for the digital pathology pathology in this case from to be used by the by the by the doctors on on the hospitals so The the other important aspect is is how you do now so that we have then to an architecture is how you Manage and do the service orchestration at at the scale right so What we are looking with with a smarter job and as well is how you do the life cycle management of those Applications and as I said at the beginning you need to look at things like reliability data privacy and and all these Different elements so what we look at from a smart job and perspective is that you can do the life cycle management in a secure way from from each of these Microservices and basically that you can connect all the different premises that are part of the of our health care Infrastructure and the other aspect will average things like One API rendering toolkit for example to to visualize the the results of one of automata Automatic analysis on digital pathology for example, so the idea is that you can get all all these different building blocks to create the end-to-end system architecture So when when you look at the Again like at the at what we are looking from from the system architecture perspective We are all also looking at the interplay within the the The premise and and the and the cloud so you will see that way if you get deeper into into what we are Providing with a smarter job, and you will see that we have connectors to the cloud service providers to a float part of the of the metadata or even like some training into the Into the into the into the cloud domain So you will see that that is it's a combination of of private public cloud on premise device Combination so that we really connect all the different pieces that that are part of our of a health care deployment so You you will as I said at the beginning we are providing reference implementations that basically are meant to Provide full implementation of specifics use cases. So if you look at this market, you will have I See this in two different ways So one that is more on the on the on the core services or in front service orchestration And the second one is on the Implementation of the use cases that you can get to one or the other right or or both so in If you go into something that is called the H over hub that we provide so you can download for example This use case on on remote patient monitoring and and and telemedicine So basically the end to end use and end to end use case that provides implementation of a stream processing from a Patient in the in the in the hospital it provides as well x-ray analysis to identify Different potential pathologies on on on the patient and it provides as well the the interaction with With with with a nurse with a doctor through through through the edge edge connectivity Similarly, you will have we you will see that we are Working on on this domain heavily in by a matter of fact now And now we're looking with colleagues like Prashant in how do we do federated learning at the scale of one of the challenges that you Will see today in the in the health care domain is that there is a lot of data out there and But again, there is a lot of sensitivity And on the data privacy right and and basically one of the challenges is for example, can we do digital twins? Probably yes, we have the right technology We have the right models. We have the right software assets, but What what is the the the frontier and how you keep like the the data? and the data privacy in You know to the to the end user right and then other things that we were heavily working on on on hardware-based security and that includes on how you can protect models that may may Identify certain pathologies how you can do private 5g deployments in in health care domains and how you can Help into the for example in the connectivity of all those devices that I was telling on at the beginning So healthcare is in general a very big area that I think it's gonna it's gonna grow over the next few years and even I'm focusing on in the horizontal edge architecture from Network and an edge division. I think that's a healthcare is a very interesting one And and it's gonna grow substantially over the over the next few years so as I said check out their in reference implementations on on a smart job and you can go into the shop or have Here today. I presented on on on health care, but we we do stuff around transportation and transportation Safety and and the other vertical so you will find multiple reference implementations that you can that you can access and use and And if you and obviously you have like a software clapper cloud as well that you can test some of the hardware that we we are developing So another reference implementation telepathology, you can download it and and you can try it out and and obviously you can ask for for support if there is any Anything that that needs to be improved or any bugs obviously we provide support to that So I think that's I'm just on on time in case you guys Have any questions if it on health care I'll I'll try to do my best to answer but in general any question on what we presented Edge any just feel free you'll have to go to the mic and ask but But that's that was for today I don't have any question good. Okay, so guess that that's it