Courtesy of SAP's Rahul Lodhe, below is a transcript of his speaking session on 'Challenges in realising AI use cases for enterprises and the way forward' to Build a Thriving Enterprise that took place at BTOES IT Infrastructure & Cloud Strategies Virtual Conference.
As we know the success rate for AI projects are very low, many use cases reach till successful PoC phase by creating ML model and desirable outcomes based on sample data; BUT moving them to production with real data starts coming from various sources ex, S3, ERP , Social media etc, and then the challenges start.
With changes in the model during training, accuracy in Inference goes down, versioning of dataset, model explain ability becomes important. This session discuss systematic approaches to take for scaling AI use cases.
It's now my pleasure to introduce you to Rob loading from S A P.
Rah. Rah! How if you could join us. It's no whatever. Like to say that rho is a world-class engineering leader with nearly 20 years experience in designing, developing and testing enterprise software products, whether with expertise in domains, like analytics, business, intelligence IOT, and machine learning. So at that point, I am going to pass you into the highly capable hands of raw whole. Rob. Thank you very much for joining us. Really looking forward to this presentation.
Excellent. Thank you.
I'd just like to share my brand.
Can you, you meet? This shouldn't variance, I don't see it!
Yes, now I have, my apologies everyone, that was my fault, not Ruffles, No problem.
OK, yep, yep. I hope you are able to see my screen and are able to hear me loud and clear.
Can you put the phone? You're able to see my screen?
We can see your screen, and we can hear you loud and clear roll.
Exactly. Thank you and welcome to the session, and thanks all for joining around the world for the for this session.
The next 30 or so minutes. I will take you through the different strategies for the for enterprise adoption. And then we also talk about the challenges that we would have to adopt the strategies. And then we follow, Let's get started.
So typically, what we see, the video of enterprise, is struggling to make the maximum value of their existing data.
What do we see is that the many enterprises help build that data in terabytes and petabytes over a period of time. But when it comes to, you don't realize the value of the real data, there's a big struggle. And the mini data science initiative, they start. So everyone wants to become, you know, next, like Google R. Over R plus L where they really use the the technologies. But the reality is far away that most of the things you do get free added when they try to go for the products are great. Solution. There's always disconnect between data, scientists put it I need versus what they want That's, you know, struggle That enterprises that enterprises to look into is BT.
The way they look, the landscapes are complicated, right. So the current landscapes are the way with Duplicate. It's being built over the ages, and I think everyone feels that they have different data stored in finance data, talk to HR data. Because the sales data, as we did different location, and we did that, also, data has been, yeah, somebody can see some things, that one should see something. So there will not be the challenge about all these different data can be, can be, not, unless they have, inside your application and liberty with the tuition. But you'd be surprised that most of the enterprise customer, may be see this disconnect.
Also, quite similar to implement the EEI, the Indian organizations, considering the, they've had a window to finish it, you can take the schedule, when, supported by the management audio. Their video, yeah, good, let's start the day.
But the moment you reach to the ...
Project leads to the product, views his taste, but getting them to real production and integrated with the realization of that, that sometimes, it's very tricky, and it's very difficult for them to, what we see, very few reaches to the. enterprise level.
And there are a few reasons to doubt about the approaches and strategy that they take, can make it happen that that's something to discuss with addiction.
The other charities to comply with data protection privacy, the company has to spend a lot more money to make the regulatory compliance, Which is pretty huge, is going to cause you don't cause our 70 on employee, because it's a lot of the issues. So be able to be careful when we talk about machine learning, and data science, and an invitation, as well, because it's a very fine gap between what is the private and personal data versus the data that you feed to model. To learn it, and we kept the IUD.
So companies has been struggling on that front, as well.
And it, the strategy that you look forward to implementing, for enterprise, has to be considered compliance approach very much. So to address these challenges, we really need to re-imagine their business. There's nothing that we can think of it that way, currently, where businesses book, I could work with other way that we have been operating for the last 10 years. I could operate in a similar way, today, could take the advantage of the AI.
So, there are, importantly, four pillar that we look forward to. How does, you know, you should look at your enterprises to be made into business versus the first and foremost important to take the data out of that, the value of your company. Theta, so, your data has been, has to be somewhere where it has, to the usable format. Your data has to be some very hard to little bit of organized in such a way that some meaning can become out of it.
It can go visualize for machine learning data sets, you have to make it as easy as it did not know beta there, but before you, you know, just to a raw data versus what output of the data.
So that we probably wouldn't making data marts are no big meetings, meaningful structure of your data. That's something that you can focus on.
Second part is the simplify and unify your data. That's something which is important.
So, you have to analyze, on correlate to the data has been there in the multiple data sources, the distributed environment. So you have to get those data to get there from the multiple data sources.
And then can happen where you can able to access the data, Because the data are available at the multiple places, is not, hopefully, where you need to be, at least, have connectivity across the data center. So, first slide that you kind of make the metadata available, second by trying to build a data and also simplify the the digital education.
Third part is looking at the how the hopefully will be confidently.
So, how the AI comes in picture is important that we are talking about machine learning models, we need to be the models of your data thought out, and then you need to really think about how you're OK.
What happened many times. It's typically, we do that.
These the AI, machine learning model building activity starts family, and then the data that we get from ideas in Excel sheets and do the searches data of any really, try to build a model and try to deploy to production. Data part is always disconnected, so because there's some data that you've got an Excel sheets on, the production system that's available for you.
So you have to really look forward to make your data which can be available confidently and you understand that how does that that particular machine learning models would connecting back to data with the Automation.
The fourth part is compliance that we spoke about. So we need to have most of the complaints. We also ensure that are there.
and, as well, the, one of the, the level, very awake and knitting the governance of the data. We can see what kind of data. How can I use the privacy of the data or the ship of OM?
Getting data is not keeping the private data grid. So, I need to have the data has been properly managed and gone to the time I could use that Also former Chief Deputy.
So, that's something we'd give me the confidence that what I'm showing is, or what I'm introducing out of getting it. I'm not violating any of these for the key pillars Plexi to re-imagine how helping it grow.
Let us go to do more in detail. So, what is the ES Apigee on the top? What Price?
So, especially promo from an AI perspective and the technical perspective.
And then with that, didn't think these three important things, like, so there's just more of technology aspects of it.
As we talk about all the transit and technology has to work together.
We'll look at how the helpless scale of scale, the AI experimentation that we're doing.
So when we talk about the scaling, artificial intelligence is more on how I'm designing, however, the blinding heat, and how am delivering across enterprises, so it's not always that, when we build something. I'd say, it has to be, look only for that particular model, a use case, but what you're building it, it has to deliver to the, sort of larger purpose of the enterprise. That's why we say, well, how am I going to deploy and deliver the EA across the different enterprises, and sophisticated, or somebody, that I'm building it for, the Division? It has to scale to multiple devices, it, for the one distributor. Or what dealer, a particular model that has to be scaled out.
The different distributed in a country or different languages? When we talk about the worldwide adoption aspect. So, you have to think about that kind of aspects. And when we are, talking about, does aspects also looked at how to scale out, in terms of a tremendous way.
So, you need to give raise to Kubernetes, are no more elastic, complimentary to technology.
And not to make it more than one monolithic stack. So, that's something we try to take it, from the scaling point of view.
How do you expect the value of your together, something, we touched upon, as well as a way to get together, and the beauty of tools and technologies to make that up.
And third and most important pillar is always you to embrace the new open source technology is not that I could build something. We planted AI is the opposite, the biggest area of a new open sources are polluted.
There are different different technologies has been the invention and we're deliberating. It's not that, you know, this all the different problems.
It is a new niche area that have been developed now, every open-source type. This also unique problem, that's, that's someone who are facing, and it's not that we, you know, we should not distinct on ...
strategy to only use whatever. I have, DVD should take care of security.
You should go, third party, open source compliance, and then, all those top 10 things, Q one, But at the same time, your architecture should be such a way that you've raised that technology as well, And that's sort of talk about strong architecture.
Little bit deep dive into the How can we scale the business, so, before that, there has been also, the notion is that AI has to be able to leave the expectation.
That we will see the good turnaround time of AI. No, I mean, that, that's a fact.
That, the fact that said it is the hype about AI, use the ....
It does science technologies, people, as the, you know, something create happening on that, does everything that that people think are imagining is getting transformed into the output?
It's not not that there are no vote. On the 64% of the demand that we're able to keep it. At the same time, there is complexity that we struggle with the Blueprint, and it's not so high.
It's not as straightforward as the ... software suite.
The same time, we also have a shortage of data scientists, or the resources, that we need to work on the ambiguity.
Last few years, I've seen every year or every month, we are getting many people getting to the causes of data science.
We have a lot of material available, and because I'm sure, these are, the shortage of that would be managed, but still, going forward, we see a lot of potential required would end at the moment and as we speak.
For the bigger point community, that's something we have to acknowledge as well, I mean the new city or challenges, but there's huge opportunities that that's upcoming as well.
So the eight core technologies, so not of the, there'll be a huge demand for the in the business was hitting re-invented using machine learning, deep learning.
There has been the scalability across the enterprise nineties, which is required, use it out of the limitations, are what we call the robotic process, automation, or religion, or permission.
So many of the jobs that people are doing, those would be done with the robots clinical and which is happening. The reality, as you see, that, pays the most.
Most are disrupting places in the industry today.
Third parties, the interactivity was going to change. So, today, we still quite depend on Alexa.
As you see the way, the interaction of you would have been changing the system, Not even with the computer system, which environment that that you look up, right? So, you already see that, as we know, many, many devices coming up, their new ....
I'm sure the video of God, as the automobile companies have been leveraging that, you use a command to open the windows, increase the volume, reduce the volume, You see million.
Alexa: I easily configurable. Obviously, you don't increase the temperature for my hours, or, so, which of these lane change the volume, change the channel. All, those are the initial stages, that we see, how the ... interact with the system that uses. Those are going to change more on the, on this. So, you have to be very cognizant about the fact, I'm going to be, not of the mind going forward of day, at the same time.
Always, we do change the future.
So, there is a there is a way of doing it. And today, the another challenge that we foresee is that there's a disconnection between the document.
So, today, we have a data scientist who wanted to be the different need for into today's landscape. You can keep it the landscape. You want to build something that the plaster phase, BRCA, all organization, who is more concerned about the cost and all those kind of stuff. But, at the same time, the IT, the IT, guys, really want to take your solution problem, your data science, to the production.
They don't understand what these guys are doing, and I'm in the mood of what about compliance, security, and all those things.
And putting things that the production, they never seen before. Not because documentation support, maybe the third parties that they use del W So, this person have been segregated ...
nine your output, the more complaints to the to your enterprises. So, how do, how do we need to re-imagine the overall business process?
And how we make it available to the enterprise AI one?
There's a percent, What does the enterprise? The, before we get into the, how do we solve this problem, but, importantly, the enterprise AI is nothing but the when you look forward for connecting to the end to end solutions. So that you Either you Or mitigate the UN data, you'll make your machine learning model.
But, AI model can only solve the problem. Not the important that you are trying to solve the business problem or not. You are a message, and connect you to the business is important.
So when you think, through this end to end, B of it, where you are, data, is available, has been curated to create machine learning model. And you connect that back to either do this application or your RFP on the chat pod. Or argue for ego analytics. So we're always taking the decision and take action on top of it.
That is connectivity is vertical support. So when you think about into any aspects of it, that for the Enterprise, AI comes into picture. one of the ways I can put my uses, the data, and I can show you what the desert is, No use and business not Solve the business problems, and that's an enterprise. It comes into picture.
And how does the intelligent information management support here, right. So, importantly, the tools that you have been using for the last decades to this tool may be the sum, which may lead now, or maybe the way data getting good editor gives you have what are the various venues to good data and store there.
Now, we have different kind of data. Right, We have data coming from streams. We're coming from social media data stored on the Hadoop kind of big data data set.
So we need to have a tools which can able to complete this deductive way that the data has to be composed together in order to build something really. Right. And then, you need to have it orchestration framework, or they don't look at the data that you get from the different data sources. And how I can orchestrate that that data.
Is it just go on prem, I'll just speak your data is coming from the cloud, that suddenly we had to comply with the governance of the data, is something which is participatory.
The Goldman scan can happen at the same time.
Didn't have either building the system, so who should see what data, how the compliance you can get, and that, hopefully, in this case, important part of the strategy is this whole architecture. Or should we make the strategy?
gotta take the form of picking the assembly line, the assembly line, which connects to your data, the assembly language, to learn things and only get scale and consume the theta?
These are the very important parameters, that that's something we have to detach upon. First and foremost, that connectivity part. So, either you are connecting to the same data, structured data, unstructured data, and all those different datasets. That's that's to take that. how does the data connection has been made there.
Once you have a data connected to the all the place this year to orchestrate them into one place, and once your data is available in ... format, importantly that, then you look at the data preparation part of that data preparation.
And to filter data, that's something that you need to know your data coming from, different sources at one place. And that data is something that's important for your machine learning model. That's something you're taking it down.
And now, you start with your machine learning model, or deep learning model, training the model, given. one model is the model.
In a little section, and then you look at, how I can scale. And so, importantly, you first, deploy the model. The deployment model is not only the bigger thing you might sound particular influences, but important part is that you should be monitoring your monitor.
So, once you monitor, you know, the accuracy of model hallways accuracy, it's providing, does it do, getting more esthetic place holders, the overall things working there. That, if I may, also get into some types of silicon. The training data is proper, or not any point of time, I could retire to model as well. So, you know, replaced with a new document, It's also the important part of the modern life cycle management and otherwise, do not get accurate results. So you need to an infrastructure where you could able to scale these dysfunctions tonight and also the order scaling, but also the vertical scaling. Stuff gets assessment.
But, the last, an important part that how that you can consume, because this model, as we spoke before, if you really wanted to make the enterprise sales strategy for AI adoption.
It's not there when you only talk about making these models are available, or they've been on the, on now, on your abusively production system, has to be sold. Some business has to be integrated with your business applications. They had to be integrated with your.
With your RFPs or origin, you know, ... group. And, that's where you say that the electrons will be complete and this is your strategy, adoption.
If you follow this versus, I'm sure, you know, how does your data scientists can village?
And you don't make the complex part, because when you look at the assembly line, where you are connecting all the previous week, the transference between the data engineer And your question was actually like the model of Like, if they did during the actor into the, what they're doing for the blind, you have to say. The same time, since, you know, all things are running and how you have to be complied with that, we can also manage your costs. So if you look at the AI is not as isolated organizer the fire, and also at the time to a part from village in stock from your data engineered, it assigns DevOps. Then also be sure you will have a different aspect, someone this tiny.
Let's look at a quick example and now you know, we just last week we look at the example of how to manufacturing companies manufacturing companies it submitted the piston. Or you can do that.
And this assembly building is typically they do manually earlier. They want to leverage the technology part of it, and technology they want to leverage via the images. So the cluster to die or images, that you apply on the particular component, the pressure.
And then you put the pressure sensors deal with it, with visual, this particular company is taking. And, again, you have data to give, you realize that, What?
we're going to be able to make, like, smart Quality App there, to be able to say that the, this particular component is a reliable, and also be paid, that how the future, currently, if you see how it could have been implemented, if you take the different technologies, that a goal. So, what resistance of the IOT and IOT Data Source Code in the Kafka Streaming?
The images is the use data, which is typically can reanalyze the region, and this data is .... And then you have some are models that something you could do, the image processing offline, to check that you know that it's been billed.
as it should you, and then you have to realize that maybe the league in ACP hana on using it. So this connection, when I don't think of my Kafka data merge with my image data, which has come, and then I could fit the data for big data analytics on Sunday and that put it. They could be like this, I know in my overall infrastructure.
But if you think from the pipelining aspects of it, or if you take from the get orchestration from the beginning, where you could make the orchestration and pipeline in such a way that the data is coming from Kafka.
And also, I'm getting the data, what I see with the image processing that I gave directly to on the Hadoop, you wrote a synthetic one particular images and the consortium that it I said, could be, Hannah, we do correlation, and then it, can you link to, that is the quality app?
Most of the production data. So this could be the pipeline that can be seen. Flow, the data analysts, and you can use functions.
And Louis De Fomento a point of view.
This is the way we feel that really good to embrace the open-source. So look at this Lego. Building an orchestra, where, you know, there has been a different, different instrument we played with the different, different players, but, at the end of it, it has to be orchestrated properly, and it should create, and, you know, good music.
That's something someone, like really, when you maybe send them inside, or they did not make sense, but when you sum them together, you will be basically, it makes them good memory.
Same logic applies here as well where, you know, you look at the data coming from different sources, should be embraced the open source so that you can spot icon. You use many new technologies coming up for renewal models. Are we not writing part of it?
Everything has to be done based on the stack which is growing but the hyperscale that's very important. OK, and then you know your private or on prem instance of it, or do you have a cloud service?
Make sure you are learning something technology, which is Docker, and Kubernetes, which is the elastic in nature, which can go on In the food sector, need It, can, it can expand and contract, so, you don't need to assign dedicated set of resources for that. And, can also be, scathingly, They can. Also, help you no indication That, quickly, done with the business processes, and we can get that, are, put it, the faster pace.
This is the way, ideally, the strategy should we look at the idea of bringing data you at any time You have a waterway which can source faster in order to connect with your customers.
So, you have a team from end to end perspective?
Um, the point of what we assist is that it started with the comprehensive information management across the enterprise, so your permission.
From them, video, any of the data or the judicious part? So once your data is set up there, you build your visualization layer, so at least you can visualize what data that you are coming from, different sources. Start doing the Michigan the instrumentation.
I use Kissinger environment started bleeding, then Attractively, so they can be used for making better use cases.
Then you look at for deployment of machine learning model, that model. So, you know, data, you have any additional setup, and then you are getting into the machine that can be directly with it, gives you start directly, pretty good received. That is, what is it?
Without having a straight stronger machine learning more than we knew anybody could be sustaining the products.
And you have to think a lot of packing, which we should Not sustainable, Right?
And thought I'd put it behind the augmented your business process automation on conversationally. On do you wanted to bring that to your analytics on your, back to be part of your sources? That's important.
So that's the way we didn't cut that.
We've been doing, one, realize what a citizen, the price, they went through the other two things.
With that, I would end my presentation and open for questions.
Quite good. Doing good or bad.
I think you're doing perfectly well and it's eyebrow. If I could ask you to, stop sharing your presentation, that would be fantastic.
Thank you once again, for a master class in your presentation. Very much appreciated, and I know that from seeing some of the questions that are coming in, that that's very true. Rahal, if I may, starts simply by asking you a very simple question for you, Not for the rest of this, of course.
AI is evolving fast and the topics and technologies are changing at such a rapid price these days.
How, how can we be sure that we are taking the appropriate steps in the right amount of care, when we're looking at building enterprise architecture, What are your thoughts around that, if I may ask?
Can you please repeat the question, there was little lag.
Yeah, absolutely, not a problem. What I was saying was, AI, as you know, intimately associated with it, is evolving, topic wise, and speed wise, technology wise, on almost on an exponential basis.
When businesses are looking at what that building, from, from an enterprise architecture perspective, what should they be looking at to ensure that they're taking the right steps and putting the right sort of governance and care in place?
Yup, Pretty good question.
I mean, this is what today's most of the enterprise architectures should be thinking about profit as a point of view.
First and foremost, thing I owe to the scalability. So, you know, they had to look at the the environment, which could be scaled at the higher cost. Take the advantage of cloud.
Then bring on prem, then toolkit for how they can devise their applications. How can they go to the letters? We bring that kind of chip.
So there would be always the case where, you know, they can take a technological advantage and would they are all applications across the mosquitoes.
And also from my perspective, I should look at the, you know, the political order. Not to be have one app per scholar, and you build everything on. When I put scholar, but you get to the digital transferred to the somewhere else, you can pull data, like, good frameworks available, like Gaitan Kubernetes, which we typically use to build your application, when I first got there. That will cause to the different groups as well.
So, I'm in the cloud strategy is something is important to try to make it cloud strategy. But also, to hook to the technology, which would stop your enable me to leverage to put towards this, is below, which is B, no. Learning, being within the organization has with other enterprise organization as well, where we try to predict to do something.
And then, you know, it just did a two year time, everything, that the whole game got change and that they do look for a substitute for, for that and then sort of tells you where you have to know something, which is which is very, very tight integration there. So I didn't try to be as generic as A, as you wouldn't be possible that you will generate component in your pictures, but, but try to be where it says credibility and something like the. Architecture, which allow you to move across to the, to the ... going forward.
I think it's going to be very common ways there, you will, how common I picture for managing the multi cloud systems, very easy. So multi cloud is distributed. Starting to that.
Automation, should we think about this? right.
Thank you for that Ronald there. That's very much appreciated, Great insight. I'm moving on to the sort of next one. I suppose this one's slightly more.
Difficult, depending on people's usage, around. Use it. I mean, a lot of people talk about it. They don't necessarily know, Especially if you're in the business side of things like myself, You don't really get the insight that you should do sometimes. But how do you integrate your business processes?
With machine learning, are there any sort of real use cases out there that are really showing best practice? Because I hear a lot about people saying that we, we are where we're going down the route of machine learning. And then you say to them, well, what does that mean for your business? And they're like, well, we're going to have the reach of machine learning. I'm so And I'm like OK, but: what are you looking to get out of it? What should a business be looking to get out of integrating machine learning with their business processes?
Yeah, that's that's that's the reality. That's what a deadweight don't even want to go with the machine learning and leverage. But.
It's more a buzz word, but businesses do, I think they should be very clear about what, what they want to achieve, the district to start with the use case, that they have in mind.
Then also, maybe you can take a simple example, something we built for the for the bank, right, and the credit risk was the challenge that they wanted to see Can you give me a little bit on the right, the liquidity or How can I get it is for a particular purpose the particular planet now, and then what are the parameters? Should I? Should I look at it from that possibly?
And we started working backward from there to see what our data available.
How can we see these guys with me, know, all the places, and then we started with ticketing based on that.
So the important part is that, what problem you're trying to solve, the technologies, they were going to be Chinese technologies, is widely available for all the things. And you said, always, you will be there, let's say, five to soil. And then you come back to the technology stack point of whether they are possible.
Do we start with the other way around, looking at what data? And what outcome I can get. The next type of retrofit that is as follows: That they want to look at the total historical picking your showing that you don't ... machine. Let me. You'll never sit here. You're trying to solve the real business phone, answer out of five fold, you couldn't get to problems like the problem solved. But at least you are solving something which helping the business to make money are helping with this, reduce the money in some cases. But you could try anatomy around things like and you're very strict guidelines As well. Whatever you're building, it has to retrofit where business process.
Fantastic. Thanks. All there.
Any sort of online, sort of resources or case studies that you would recommend directing business people to go and look at, before they start going down the route solve, really saying, this is what we want to do. But to get a better handle of people who've been through the journey and really know what they're doing other than, of course, knocking on your door. and saying, well, can you come and help? Which is always, always a way to go, but any, any sort of resources out there, which we can access to really, sort of make sure that we're, we're saying the right things?
Yeah, I think this Kaggle is the very good article.
Is this part of the Kaggle website that you see, I think there have been open source, and many contributions have put into that, as well.
We also, they sometimes do that, that list to see what are the ... and contributing to it. So I think that's something that would be the good, good starting point.
Decided to bring the resources where you would see the real time examples has been published in different ways to solve the business problems, and they also some test data available, or something like that where you can play around as well.
So, that's where people can get started, too, See some kind of meaningful outcome.
Fantastic. And just got time for one for what may be 1 or 2 more questions at this stage.
one of the big things, the, especially from a European perspective, but it's it's going to, and is increasingly drifting across the Atlantic and father globally, is the issues around data protection.
How you address them, when you are designing an enterprise architecture for AI, is obviously quite an important part for a lot of businesses now. The penalties are extreme if you get this wrong.
And can affect the ongoing health of every business, if you get it massively. What stage do you build the scene, But how do you address it, is a straight at the start as part of the kind of strategic road mapping? Or do you bring it in once you've kind of got the other parts? From your perspective, where is it best to sit this?
So, it's a very good question, and pretty, you know, near to me as well as the build a new application. So we are in the process of building some new applications, as well, and the point. And then, also, what we built in the past been, as you rightly say, that a lot of ...
coming up, I only mentioned privacy. What should be seen? What should not be seen? And how you design your yourself period.
Uh, the later part you try to do, your theory will do it.
So, you need to have that. I had to redesign.
The moment you start thinking about your application, that that Network Especially maybe little technical bill for particular crosses the, when we build, it might get an application. Right. So, typically nowadays publications or get to the cloud and then the dependency and goals. And how does your current setup account.
You also create a new concept called zone of bullets what, or who can see what kind of aspects there.
So, very meticulously designed your, your software, considering that, who should see what, and what revised as he should.
So, yeah, baby, rebuilt the machine. Learning's which data is very important.
Cause We see that you model to the cloud. You bring your data training, training, your model on the cloud.
And then we train there and then you get your inferences or your output of your your time data. Once you do the training.
Now we have to be very careful when we ask for training the data. Because me as A ASAP or B, as a as a as a vendor, I shouldn't be seeing the data of our customers. So it should be breathing indices.
Dinner space is going to use the cloud account.
And when the training done. and so the liquid activity for running his model accurately looking for data.
Very, very well, made my division of what data coming in and what I've been processing of the data.
So, processing and the data logic, but you have to differentiate and at every layer layer and be realistic in a European company. We are very strict about this is all the norms, and for us, when we start designing the picture, first thing we look at that, OK, what data protection privacy could come? And what is the ticket come to? Where ticket to be able to contact customer? And they'd only be able to look at the data for the ... audits ID in our cloud workload, but we should know what should be allowed.
We'll go look at the data.
Same thing happens when you are interested in audio. building that back to digital business services.
Who should see what is very important? So, these are the questions that you have to answer at every stage. Does this person is authorized to see this data, especially when you're building a cloud of division within the, within the organization just as well? Or when you're building the bigger product, like the what is it will enter into big problems?
Data Security has to be taken at the first level of your picture part, and your architecture nowadays, with the protection privacy, and data security aspects of it.
Then you build an application and the bus basically applications, we're also, yeah, I've been, I think the team without considering this fact and we said, OK, let's get rid of that application. Available bandwidth to visit stack is very difficult.
Change the nuts and bolts of times I would say consider this as the first requirement as we start building the right picture.
Fantastic role, thank you for that. And thank you for sharing your knowledge, so generously asparagus. The time to join us today. Very much appreciated and look forward to catching up, EG, course, but for now, thank you very much for your time and input, I have a fairly enjoyable rest. Awake.
Yeah, it seems to me it's always a pleasure to talk to you, Ben, and presented in your performances.
Fantastic. Thank you for your kind words row. Take care.
So, everyone now at this stage, normally I'd be saying please rejoin his again at the top of the hour for the final session in the conference series. Unfortunately, I've finished form that our final speaker, I call it join us, today, which is a shame. But, as a fantastic presentation, we've been really great, However, I hope, that the previous sessions five worst, where the appetite to explore and engage around, your infrastructure, and what you can do, now, that the cloud is really tangible part of what we're doing within the business community.
These sessions will be available on demand, as other sessions in this series being very shortly, so please feel free to access the links that you'll be provided to join sessions. Any queries or questions, please keep the conversations going online, either through my colleague DJ's link, which you will have seen coming up throughout the conference, or with Joseph Link as well.
And finally, as you may be aware, we have our flagship vetoes for home virtual conference coming out from the 17th to the 19th of November, which promises to be the most interactive cross industry event of the year, especially within the transformation and operational excellence space. I can promise you that you won't attend another event like that this year, or maybe for quite some time. Lots of opportunities to interact with P is Learn from solution providers. Contract discuss meets 1 or 2 unusual sort of nuances engaged as well. So finally, thank you so much for spending the time to join us at this conference. I hope it's been as enjoyable for you as it has ferocity presenting it to you and as I say, if there's anything else you need, please let us know.
All it remains for me to say is Please keep siphoned Well and thank you. Bye-bye.
Senior Director SAP Artificial Intelligence,
Engineering Leader with more than 18 years in world class company having extensive experience in designing, developing and testing enterprise software products. Expertise in domains like Analytics & Business Intelligence, IoT and Machine Learning.
Responsible & accountable for driving global missions with Product Development, Processes and People. Managed end-to-end life cycles of onPrem and cloud enterprise software product components – from whiteboard designs to maintenance releases.
Experienced in building ground-up world-class teams, as well as in driving transformations of inherited teams. Building the team with Innovation as DNA and transformed employee engagement and trust score. Extensive experience in working with geographically distributed teams. Managing a development team of Managers, Architects, Product owners and Program managers.
Excellent balance of Technical skills, Managerial skills and Entrepreneurial skills.
Thrives in a 'start-up like' work environment.
Search for anything
November 9, 2021
11:00 AM - 12:00 PM ET
January 13, 2022
1:00 PM - 2:00 PM ET
January 27, 2022
1:00 PM - 2:00 PM ET
Watch On-Demand Recording - Access all sessions from progressive thought leaders free of charge from our industry leading virtual conferences.Watch On-Demand Recordings For Free
Courtesy of DC Government's Ernest Chrappah, below is a transcript of his speaking session on 'Going Digital To Enhance The Customer Experience' to ...
Courtesy of 's Anu Senan, below is a transcript of his speaking session on '' to Build a Thriving Enterprise that took place at Enterprise ...
Courtesy of Tasktop's Dr. Mik Kersten, below is a transcript of his speaking session on 'Project to Product: Driving Digital Transformation Insights ...
Courtesy of Nintex Pty's Paul Hsu, below is a transcript of his speaking session on 'Improve employee productivity during and post-COVID by ...