 So, good afternoon, welcome everybody to this session. My name is Margaret Laughin and I'm the Business Development Director for Machine Learning with SAP. So, my day-to-day job is talking to our customers around our strategy, what it is that we're doing in machine learning, the solutions that we have coming to market, and of course the value that we're adding here to our customers. Now, because I know this is a technical conference and a technical audience, I'm going to be joined today by Tejin, if you want to stand up and say hi. And Tejin and I are going to tag team today. I'm going to cover the business side and talk around the business value that we have with machine learning and what we can do with this. And Tejin is going to talk about the technical part, so how we're leveraging Cloud Foundry in the applications that we're developing and all that good stuff. And we'll talk a bit about the API Business Hub and if you missed a session, two of our colleagues here, and shout out to them, did a session previous to this around the API Business Hub and how you can leverage that, that's going to be useful to know because our machine learning functional services are available through that. But as I get started and as we start to jump into this, hands up, who's familiar with what SAP is doing with machine learning currently? Okay, they're all the SAP hands. If you're not part of SAP, hands up if you're familiar with our strategy and with what we're doing. Okay, that gives me a wonderful opportunity to then get started on this slide. So yesterday if you were here and we had Bjorn Gurke, who's the CTO for SAP, he was giving the keynote. And what he was talking about during that, of course, is digital transformation and how this, how our customers are looking to us to help them enable, digitalize across their enterprises. So how do they connect all the different people, things and processes that says that they're in the side? And of course, what are the different components, parts of this? So one of the things we know when we're talking to our customers is that this is quite complex for them. And this is very challenging and hard to do. The other, that's on one side. On the other side as well, is if you have a resource budget or a strategy budget and you're looking at innovation, of course you want to make sure that you're doing the right type of innovation. Because innovation can be very costly to implement, and innovation can be costly as well in terms of stakeholder engagement and getting folks on board. So with SAP, we can bring this together across digital transformation. We brought out SAP Leonardo, which we launched back in May during Sapphire, which is our leading sales conference. And what this is, is a collection of technologies, as you can see here. And what we're doing is helping our customers across their value chain, either help on the business process side or help them innovate in many different areas. So the intelligent part here, of course, is the machine learning part. So I'll jump into that next. OK, so in terms of how we apply it as well to the enterprise, I mean, we're all familiar today with machine learning under consumer space. So I'm looking around. I know everybody's got their iPhones. They're looking at your Google Maps. You're getting up the time for traffic. We've got, you know, well, HomePod just announced there's Alexa. So there's many different types of tools that we can use that help us on the consumer side. But now on the enterprise side, we're seeing machine learning really enter this space. And of course, there's three things that are enabling all of that. And they do this on the consumer side too. But of course, we know that our infrastructure is better than it's ever been before. So leveraging GPUs and folks like NVIDIA who do great things in this area certainly help us take the power and the capacity from what we have in terms of the big data and what's available out there in the masses of data that we're getting either through our sensors with IoT or we're looking at our business networks or across different landscapes and how we bring this together. And of course, sorry, this is where I should learn to take a pause before we continue. But the third part here that we're looking at as well are deep learning algorithms. So we know that these are far more sophisticated than they've ever been before. And there's great companies out there doing super things in this space. The other part I would say and certainly for us, when us as SAP, when we're working with our customers, they want us to help them generate deeper insights into either their existing business models or help them on their future roadmaps as well. So certainly from an enterprise perspective, we definitely see that machine learning, we're in the right environment for machine learning to thrive and start to build and develop out from here. So with that, and when we talk about SAP and we think about us as a company, I mean, we're across 25 different industries. We've over 12 LOBs. So the breadth and depth of experience, I would say, that we bring into the market here and understanding our customer's business processes really sets us up in terms of understanding the value that we can add to them. So I talked earlier here about this in the sense of when we think about what we're doing in terms of our strategy and our roadmap and approach, we're looking at two things. One, we're looking at business processes and where we can automate like highly transactional activities and help our customers manage that differently. And of course, that leads to things like increased revenue or less we drive down the TCO and it changes up as well what people are doing in their work. That's on the one side. On the second hand, if you think about this, we can leverage Leonardo. So what we were talking about earlier in terms of the digital transformation view, and we can leverage Leonardo in order to help our customers productise, or develop or innovate in many different parts of their business, leveraging the information that we're getting across that value chain that was presented. So taking all of these different parts, how you can consume machine learning, and as I said, we'll start to get over into the more technical part of this shortly. But the foundation for it all from SAP is, of course, the SAP Cloud Platform. And there's many of my colleagues who are here in the room that can help answer some questions on that as well if we want to spend some time in it. But from the Cloud Platform, there's a few different things you can do. You can leverage. We have business services, so we are enabling part of our product portfolio with intelligent applications, and we're making them more intelligent as such. So there are products that we are bringing to market that will have intelligent capabilities, so machine learning within them. So that's one thing there. The second part is functional services. So this is where you could leverage image, detection, text, tabular, things like this in order to develop applications that you want. So that's where you see the top part as well. So you can build as a partner or customer. You can build your own machine learning applications, leveraging our functional services via the API business sub that you guys talked about earlier as well. So in essence, that's some part of how you can consume machine learning with SAP. We have the Lifecycle Management Hub here. So what we're bringing to market in this slide as well is trained and developed machine learning models and algorithms that you can also leverage with your business, depending on what it is that you want to do. So for those of you who are familiar with machine learning, or you're implementing this in your business, or this is an area where you want to develop, of course, with machine learning, it starts off the first question you ask is, what is it that you want to do? And when you've established what it is that you want to do, you can work backwards then in terms of understanding the best way to build that out and to drive your business. So here's another picture that gives a sense of the breadth and scope that SAP is developing in this area and what we're bringing to market. And of course, you can see how across your entire value chain, you can leverage machine learning. And it can do many different things for you as well. So that's a sense of some part of where we're going. And then here in this slide as well, we have a number of solutions that we've brought to market that work in both the back office and, of course, the front office as well. So we've many different customer stories. And typically, if this was a longer session for me, I'd talk a lot about our POCs and what we've been learning with our customers as we've brought and developed and brought this to market. But I know that my time is a bit short, and I have to hand it over to Tagen. But before I do that, are there any questions right now in the sense from the business side or the approach that you may have in terms of thinking about machine learning for your organization? Yeah, so think about Leonardo as a portfolio of technologies. Which, just go back one slide. Okay, so Leonardo, it's a portfolio of technologies, and that's that you can leverage them in different ways. Or, you know, one perfect example, I had a customer talking just recently where they have like 50,000 sensors out in the field. It's an oil and gas company. They want to know how they can apply machine learning in order to enable different insights from that information that they're gathering. How familiar are people with machine learning the topic itself and the technology? Okay, so a variety of hands. And I guess it's certainly a trending area so we know people are getting far more interested in it. A few other things I'd say about machine learning then as well as there's no one size fits all. There's many different types of machine learning and we're leveraging different technologies in order to build out the capabilities that you want to do. So, there's a different slice and dice involved with all of those as well. So I guess, you know, this is some of what we've been learning as well as we've been building out these solutions and bringing them to market. Sales and service. So one of the use cases that we were looking on here was all around the sales cycle in terms of getting ahead of qualifying your leads and the next level of sales, potential sales opportunities that you would have but it drives it a bit deeper into understanding and connecting through the omnichannel engagement and digital footprint of the customers in terms of bringing more intelligence to understanding the direction that they're going in and potential opportunities that they're looking at so you can build out specific campaigns to target that or have your sales team work with them differently. I'd have to look back at some of what we have already because I know we did it a short while ago but I can get that information. So I will move it over to Teijin who I know is gonna talk more around the technology as well but and how they're developing this within our innovation center network groups. So Teijin. Hi. Yeah, I'm Teijin and I'm working at SAP as a data scientist. So today, before going into my presentation I would like to ask how many of you have heard of deep learning? Nice, yeah. So I know like deep learning is the hottest topic in even in machine learning area and a lot of research institutes are doing research on deep learning. So let's look into how we can use deep learning in an application. So there are three major areas where we can use machine learning, a deep learning especially. So for, first of all, it's a speech recognition and the second is natural language processing and the third is computer vision. So for example, for each area is like I believe you guys are familiar with Siri. So Siri is using a deep learning technique for speech recognition and Echo is also using, Amazon's Echo also use a deep learning technique and Google translator or a lot of chatbots on the online are all using natural language processing using deep learning. And also a self-driving car or like face recognition, they're all using deep learning for computer vision technology. So let's look into their data format and their required computational power. So for speech recognition, the input data format is a voice and the voice can be represented in one dimensional array. So compared to other data set, the data size is pretty small. Like and when you look into their benchmark data set it's like 60 MB megabytes. And for natural language processing, the text data itself is pretty small. However, when you feed those data into deep learning model, you should change each word into factor representation. So for example here, like wait for the video. Each word will be like 200 or 300 length array. Like floating like 32 bit or 32 bit floating array. So one sentence will be two dimensional array. And when we look into like those benchmark data set it's like one GB. So which is like 10 time or 100 time bigger than voice data. And for computer vision part, the data set is much, much huge. So for computer vision, the input data is image or video. So for image, usually it's three dimensional array. And the size of array itself is much bigger than this sentence or text array. So one picture is usually like 500 kilobytes and the benchmark data set is like 200 GB. So for when we look into the computational requirement, there are a lot of CPU recognition companies which using just using CPU for CPU recognition because the data set is small. So there's a, we don't necessarily use GPU. For natural language processing, it's highly recommended to use GPU because based on my experience, when comparing the performance of current GPU with eight core CPU, GPU is like 10 times faster than eight core CPU. And for computer vision, we definitely need to use GPU because usually current GPU is 100 times faster than CPU. So if you provide like real-time application, even though we use GPU, if we use very, very like complicated architecture, like neural architecture model, then it takes like one seconds or hundreds of milliseconds. If you use CPU, like it takes more than one minute to process just one request. And like considering there are like tens of, like hundreds of requests will come in in one second. So we definitely need to use multiple GPU to create a good application. So inside SAP, we developed a computer vision application. So as I said before, we definitely need to use GPU for computer vision application. I know that currently Cloud Foundry does not provide GPU device. However, we can easily integrate external GPU with Cloud Foundry application. So inside SAP, we have SAP motion learning platform where as a data scientist, I can train my model and also deploy my model and make it as a service. So after making this motion learning service, we can integrate with application in Cloud Foundry. For other cloud platform, the reason that I mentioned is I also try like some commercial application, cloud platform which provides GPU such as AWS or Google Cloud. They all very well with Cloud Foundry application. So and for other parts, I believe you guys already know, we can leverage a lot of services which are provided at Cloud Foundry. So let's look into a little more detail on our application. So for machine learning, especially deep learning application, I think scalability is very important. So we adopted like queuing system and worker application in Cloud Foundry to handle machine learning jobs in parallel. So these worker application is very easy to scale up. Like we can just scale up 200 or 200 in very short moment. So all the input data will be accumulated in queue and then workers gonna capture one jobs one by one and then put those requested to external GPU machine through REST API or RPC. And another important part here is we should have security module because if there's no security module, then like other application can easily access our like deep learning model without any authentication. So there should be whenever workers send job to external GPU, they should use security module. And currently we are also looking for like other architecture as well. So if you guys after session have any advice, then you're welcome to accept that. Yeah, this is pretty much for my side. And I wanna, I think we can. So we can move it into Q&A now. I guess one of the things that Tejin and I wanted to say here as well was that, I mean, this is just like a little like baby toe into the water here on machine learning. We could do an entire track on this today in terms of looking at the challenges and opportunities across the business in terms of where you can apply machine learning, but also then looking at how you can innovate and how you leverage different types of functional services or computer vision in order to help you do different things as per like retail use cases or what have you. So as I said, it's just a kind of a baby toe into the water. So are there any specific questions that folks have for us or are areas that you want to explore with just the time that we have left here? Okay, I see a hand over there. Yeah, so, and that actually came up in the previous presentation as well in terms of how you can do that. So we have brought open, sorry, translation services to open SAP courses. So today you can access an open SAP course and if English is not your primary language or you want to have it in another one, it can translate automatically for you. So we've developed that through some of our product itself, but it's also available as I understand in the API business hub. So it's available there as a service if you want to leverage it yourself for your business. Yeah, in the technical perspective, like Google, like all the, like state of the art translator are using a team learning models, yeah. There are no more questions. Oh, there's another one here. Normally we always have hands up. There's so many things to talk about for machine learning. It's such a great topic. Yeah, it's a big topic. It's a big topic, yeah. So, yes, so, yeah, we have, we call that as a like delta training. So whenever like new customer or like new, like whenever we have to update our model, we can leverage existing model and we can train, retrain our models in very short time, like just changing some part of network, yeah. So, and I think like, yeah, we provide as a like machine learning platform. So, yeah. Is there any other questions or is there anybody working on a machine learning project now that they want to share or kind of say some of their experience through that? I think we've got a shy group. Are not enough coffee, right? Well, we're pretty good here in terms of, as I said, this was a very brief overview. I mean, if there's any things specific we want, you want to talk about afterwards, we're like, Tej and I are here. And if you have any further questions or want to explore the topic a bit further, but as I said, this is just like a baby tiptoe into the water. There's so much to the topic. There's so much in terms of what customers are looking for. And one of the other items that we, you know, areas to think here and why it's important at a tech conference to be talking about the business value to machine learning is that if there isn't a need from the customer, well then there's no need on the development side. So that's one of the things that we're very mindful of in terms of what our customers are telling us in terms of that need. So they certainly see machine learning as really supporting and helping them automate highly transactional activities on the business side, but then also how do they leverage machine learning functional services to do many things on the innovation and the development side as well with either product or they're extending their innovation strategy. Yeah, so I mean, that's a good question in the sense of and I'll say a bit of where we came from. So both Tejan and I are part of what we call the Innovation Center Network at SAP. And we have a chief innovation officer, a guy called Jürgen Mueller. So if any of you are following the blogs or what have you, you might see some of what they're developing. And our role are certainly what he, our organization's mandate is to pioneer new technologies and to see where we can build them into either growth models or pioneer new business paths forward as such. So certainly from the work that we've done with machine learning and the way we've brought use cases to market, so we've brought solutions to market. We've tried and tested these with our customers. So we've done that, but we're also implementing across our entire product portfolio. So if you're an SAP customer or you're interested in SAP or you want to engage more in the enterprise side, what you'll start to see from our product portfolio is that there'll be a huge number of offerings where we embed intelligence into either existing applications or it'll be on the roadmap for future applications. Certainly I can say, because we're on record to say this, but we're betting big with machine learning. There's no doubt about it. I mean, at every part across the, you can make your enterprise intelligent. There's no doubt about that at every part. And that's why we showed the value chain slide further along because you can see how machine learning can impact every part of your business. And it can help it just do very different things. And with the Leonardo umbrella as well, where we have that collection of technologies coming together, we're also engaging with, you know, what we have on the IoT side or big data or analytics or what have you. So certainly you're going to see a lot more from us and we're working with our customers right now on some really exciting use cases where they're helping to offer better value to their customers. So things like the retail shopping experience and leveraging things like trend analytics to, we had a use case recently out of China where they've implemented a whole new shopping experience whereby the consumer designs and models their own shoe and a week later they can come and collect it because the supply chain is also being impacted to be much faster efficient at no additional costs to the vendors there in that case as well. So that's kind of an easy example to run to but there's plenty of those that we have that we're working with customers on. I think we're all right. Oh, it worked. I can talk for a long time on this topic. Okay, with that, we'll wrap up. If you have any questions, definitely let us know and we'd be more than happy to talk. Thank you. Thank you. Thank you.