Optimize industrial operations with AWS IoT

How's everybody doing today? I need more energy. How's everybody doing today? Good, good at AWS IO.

We usually start our customer presentation by asking ourselves and our customers this simple question that is if you knew the state of everything and you could reason on top of that data, what problems would you solve? The key thing to notice here is knowing the state of everything in your company and operations and to be able to reason on top of that data, it's our mission at AWS IOT to help our customers accomplish just that.

Hello everybody. My name is Pravin Raph. I lead industrial IOT and analytics. Welcome to our presentation today. I'm also joined by two of our esteemed guests, Rohit Talwar from Penske and Rajesh Dev Nani from Hitachi. We're going to go through an introduction when they come for a presentation, but we have certainly some very exciting stories.

So it is with this mindset of really co-innovating with our customers and really, you know, helping them solve their operational problems that we have been able to win the trust of all these major customers that you can see. And we also have a very rich set of partners who kind of help us deliver on those things. You would hear from one of such partners, Hitachi on how we are kind of collaborating to help our customers to win many of the IOT workloads. And you'll also hear more stories about Penske on how they were able to benefit from AWS IOT services and solutions that we have.

So when we say optimize industrial operations, what we really mean by that is basically customer comes to us because they want to lower their operational cost and increase their revenue and also improve their sustainability. In other words, we're really trying to help our customers to make their products faster, cheaper and better, right?

So we see three broad areas where customers come to us for help with, you know, innovating and lowering the cost. One is in the factory, according to the latest study from McKinsey manufacturing in general, produces about 1800 petabytes of data, which is twice more than the next nearest industry. And the 90% of the data is locked in the on-prem. So if you want to save cost in your operation, what's the first thing you need to do? Move the data to the cloud, right? Because you don't have to maintain all the infrastructure and all.

Another thing is liberate that data and democratize the data so that different constituents can get meaningful insights to help them with their decisions in the operations, whether it is understanding the machine availability, OEE quality uptime and, and many other things, right? Because you want insights to be able to make better decisions in your operations and then moving on.

Other major area we see is in the products they make. So there is more than 13 billion connected products that are already there and equipments. So more and more equipments and products are getting instrumented. That means products are getting smarter, they're getting more connected. And also other thing that we are seeing from a trend perspective is customers who come to us also want to understand how their customers are using the product. That means they want to get real insights on how their products are used. So that way they can take those insights and then take it back into their product design. That's the third area.

So make those products more customer friendly, more manufacturing friendly, designed for manufacturing and then also make them more efficient. So which in turn feeds into the factory. As you can see the products are becoming more intelligent, you know, definitely more efficient, more sustainable. So as you can see there is a flywheel effect, so this keeps going and then this is kind of how you can continue to optimize and then extract more value from your industrial data.

So moving on. Where do we see customers see challenges and, and also opportunities for improvement, right? So first and foremost thing where we see is lack of visibility into operations, what it means really like we started lack of visibility means not understanding the state of things within their company, right?

So what it leads to a lot of reactionary measures in their operations and a lot of manual work, a lot of redundancies. So that's something that, you know, certainly an area to come in and then improve, right?

The second area where we would see is unplanned, breakdowns if you talk to almost all the CFOs in a company. Their biggest thing that keeps them awake in the night is they go through the budgeting cycle, they budget and then suddenly something happens now they got to go find funds to address that problem and not only that and also impact it would have on the revenue.

So that's kind of where they need to have a more proactive approach, more approach in which they understand, not only the state, but also reason on top of it. So that move break, you know, move from unplanned to more planned maintenance and pay as you go model. So that way they don't run into disruptions, not only in operation but also in the finances as well.

So third area is simply the waste that happens, it could be waste because of the scrap products that they make are the products that accumulate additional operations once they are deemed scrap, right? Or it could even be the recalls that they have to go through and associated brand images. The impact that would have on the fourth area where we see a lot of industrial challenges is the energy efficiency, especially in certain geos like EM, where you have geopolitical issues with the Ukraine war. And certainly manufacturers are, you know, struggling to keep up with the energy cost and continue to run their operations.

So they are looking for ways to how can I do more with less or more with the same in terms of the energy? So, so given all these challenges and opportunities, where do you get started? Right, in terms of optimization, I think the first and foremost where we recommend is getting a good handle on your industrial data.

What you're seeing here is a classical pyramid or ISA 95. However, the way you want to call it where organizationally data is organized in stacks and they are measured in 0 to 4 levels. And also it's categorized into OT data and IT data. This is great, you know, and it started with Industry 4.0. But the problem with that is average age of the industrial machinery is about 25 years and there is 300 plus protocols, they don't always talk to each other and all these layers are in their own silo.

So there is not much connectivity or the data exchange happens among the layers. So that leads to a lot of inefficiencies and even the innovation that happens happens in silo. So what we are seeing with the ubiquity of cloud, so that barrier is breaking right over the last couple of years OT and ITs are coming together.

So, so the data is not stuck in one spot like it used to. And there is also newer protocols like OPC UA that is becoming more popular where you can translate from one to another. So there is flexibility that is creating more efficiency that would otherwise happen in a IOT, right? As we go into the future, what we see is again with the cloud presence and the ability to move data from not just from edge to cloud but also cloud to edge.

We expect that barrier to be completely gone. And so that there won't be much difference in the way things operate between the cloud and the edge. So some of the newer things like EIL and even digital twin, so you should be able to operate more at the edge and make decisions at the edge even when there is no internet connectivity as though you do it on the cloud.

So this is something that we are super excited about. This trend is going to continue to evolve and it's going to unlock a whole bunch of efficiency like it has never been seen before.

From the use case perspective and the outcomes, there are five areas where our customers are maximizing the benefits of IOT and EML and cloud. The first area is unlocking the higher operational efficiencies. What we mean by that is essentially getting to know not only the state and all. So the reason, so that means remote health monitoring, like your OEE dashboards, those are all a major outcomes that we are seeing.

In addition to that, if you remember, I talked about like 90% of the data being on-prem. So this is a great area for people to move that data into the cloud. And then not only get insights at a factory but also across the factories, right? So that's kind of where we almost call this historian and historian plus.

So there are many things that you can do once you have the data in the cloud and contextualize it with other forms of data, MES your CMMM your metrology data. It it opens up a whole pandora of opportunities to uncover efficiencies like it has never been before. And then we talked about the downtime.

So this is kind of where having a good handle on the conditions of the assets and then going through the journey of like getting descriptive predictive and then prescriptive insights on what's the optimal level to operate your factories and machines and other equipment. So that's one area we're saying.

And also with the advent of digital twin, more and more operational twins are being produced. So that way you can remotely understand exactly where the problem is and also what you can do about them, right?

So quality is another area where we are seeing customers are deriving a lot of benefits, especially with the advent of computer vision. And when IO meets AI scenario, so there is a lot of things that we could do that could never been done before in terms of identifying the anomalies, clustering good versus bad and and really understanding why things work one way and why it doesn't. And comparing across different machines, factories and enterprises, right to reduce all the scrap and other unwanted cost.

Fourth area is around the energy usage, sustainability. Getting a good handle on the energy and then metering it and then getting really efficient, you know, in all forms of operation, your equipment, in your buildings everywhere, right? Wherever you are using the energy. That's a key area.

So where we are seeing and also an impact of environmental factors on your energy usage, that's another critical area where we can input weather reports into your AIML models and unlock more benefits.

Last area, manufacturing and supply chain are like, you know, born twins, one cannot exist without the other, any disruption that happens in the supply chain invariably affects the manufacturing. So this is kind of where we are seeing a lot of our customers that are track and trace solutions control towers.

We announced a solution service yesterday, AWS Supply Chain. So that's kind of where our focus is and our customers are benefiting from some of our services now.

Just to kind of give you a preview on what are some of the IOT and ML services that we have that helps you with delivering all these outcomes. We classify our IOT services into three broad areas.

One is what you can do at the device level. What is the cloud company doing at the device levels you task, right? But if you think about it, you know, getting a good handle on the device, helps you understand the state of the device and also helps you connect that device to the cloud in a most efficient way and ensure that it is secure and safe at the device level, right?

And the second thing is around the connectivity and control this matters because this is kind of where we can help our customer orchestrate a wide variety and wide number of devices and also on board them a scale, a scale is the a key word here, right? Because that's kind of where you get the benefit of ensuring that you keep an enterprise up and running in a most efficient way.

Third area is around the analytics and streaming services. This is where you take the data and you reason on top of that data and and get insights to help with all your operational decision support and also to orchestrate and automate where you can.

So what we are also seeing is these three are not layered. As you can see, I didn't put them on top of a pyramid or as a cake form, I put them in a circle here. So the reason being one feeds off the other. So as you start leveraging this is a scenario where a hole is larger than the sum apart. So the benefit you get is far exceeds using them individually. That's why we call this a virtuous cycle. So it's a self fulfilling cycle where more you use it, more benefits you would get um just to do it quickly to double click on these things.

Um we have uh basically these are the services that you see. We have four services in the device software areas that helps you connect and secure your devices and end points. We have green grass that can bring in some of the cloud capabilities to the edge if you have stronger, you know computers at computation at the edge like plcs.

And then on the uh connectivity and control services, we have five services iot core is by far our most popular service. Uh we have uh last i checked was 1.6 billion end point. And we have uh we get about 2.2 and a quarter uh connections per minute coming into us. So this is huge and this can clearly shown and proven in the industry for several years now that it can scale and handle at the capacity that is requested by our customers.

And then we also have fleetwise to help with large, managing large number of fleets, particularly automotive and the roborunner in terms of orchestrating robots in a distribution center or manufacturer or any other place that you want to kind of orchestrate your robots, right?

Moving on to analytics and streaming services. Um so this is kind of where uh we have five services uh iit events, you can customize, you can write rules and make actions on certain events, whatever events happens that you gather from your sensors or it devices, right?

Uh and then we have sitewise which is an industrial grade um software service that is built to collect connect uh contextualize data, you can build asset hierarchies. Um and you can also store time series data and make decisions on the remote health of your equipment and you can also build other dashboards to uh get additional insights into your operations.

One of twinmaker is one of our newest service uh that essentially helps you build operational digital twin. We also have partners such as matter board who can give you a 3d visualization of your operations. And uh you know, uh you can deliver some very interesting use cases and outcomes by leveraging digital twin.

Um and we also to complement iotiot and a iml go hand in hand almost if one doesn't exist without the other, we also have uh I'm gonna touch on four services here. One is look out for vision. Um so there is certainly a demos in the uh industrial tent. You guys can go check it out where you can train uh a iml models to look for certain anomalies either in your parts uh or in your operations. You can train them to uh essentially give you early warnings uh whether it's a quality or any defect. Um you can automate them as well, right. That's another thing that you would notice.

We also have look out for equipment which can uh help you with condition based monitoring, look for root causes of the disruptions in your uh operation. And we also have monitron. Monitron is very easy to use. You can even buy it from amazon.com. It comes with both sensor and the insights, the dashboard that you can and also a gateway, it can detect your vibrations your temperature. And then even if you don't have a plc, you can literally stick it on any rotating equipment and get insights on the inner workings of the health of that equipment. Certainly you can use it in wide variety of use cases.

And lastly, we also have uh aws panorama, which is basically a computer vision based. It can turn any cameras into more of an intelligent camera and you can use this in wide variety of users, whether it's in a security safety worker, safety situation, um or um essentially as you are seeing in the thing, you know, managing the fleet of trucks. So there is a wide variety of use cases wherever you want to turn your cameras into more intelligent cameras and make decisions based on that.

And then uh I won't drain you into the details. Here, I showed you the list of customers. Uh here is all the things that we hear back from our customers on how they are benefiting from our uh solutions and our partner solutions as well, right? Many of these services gets embedded into our partners. Um and then essentially make them even better. As you can see, there is a wide variety of benefits in wide number of you cases. So certainly we have wide variety of partners who have taken these things and then delivered value at even higher.

In fact, we have one here, hitachi and tara who have already taken their existing industry ready solution, which has been in existence for very many years and embedded our services and made them even better.

Um so with that, i would invite rajesh dev nani from hitachi to come in and tell us a story about what they are doing with our services and how they are benefiting.

Thanks a lot, pravin.

Thank you, sir.

Hi, good afternoon. All this is rajesh danni from hitachi var app. And uh hitachi var app is the global solutions services and digital infrastructure arm of the hitachi group from a global standpoint.

So pravin, the last light rin presented, made me recall and remember a blog post. I wrote about five years back which was caught putting the iot cart before the horse, things have moved on from there. I think it used to be all about technology at that and all the bells and whistles of iot, the focus is really pivoted towards the business benefits and the value of that. And that's a sea change from what we saw before.

So look at it past uh now in the current context, five years, a lot of ch has happened. A lot of iot platforms came and went and only the ones that are focused on the real business value have stood the test of time really effectively. And a w is obviously clearly being in charge of that.

What i'll cover in the next couple of minutes is uh i'm gonna give you a brief walk through in terms of um hitachi's uh credentials and vision strategy in the iot space and how we are aligning with the a w who's again the market leader in this space, but we are saving the best for the last uh so we'll have rohit come in after me and talk about how we are really realizing that in a customer context and working in a collaborative journey with them to make it all happen.

So just to get uh started a bit of a brief on hitachi, uh i'm not sure how many of you know about hitachi, but uh we started life about 100 and 12 years back in a small uh workshop in the prefecture of uh hitachi. And that's where the name comes from basically.

And uh from those humble beginnings, we have really come a long way. So it's about $85 billion in terms of size now. But what has really stood constant over all these years is the spirit of innovation that parated us right from the start and it was all about creating a positive societal impact. So that's really stood the test of time as we have grown in, hitachi, hitachi has grown to become an industrial giant.

So we do have real credentials in the industrial i space. We operate about 200 manufacturing plants globally across the board. Uh we do work on all the critical infrastructure sectors and we do provide very complex high value assets. Not only do we design them, build them, manage them, monitor them. So it has been kind of in our roots, the whole industrial thing has been in our roots essentially.

And we also have a digital practice which has been in existence for about the past 60 years, the it the it side of things. So what really accords this is gives us a very unique position in terms of understanding both sides of the world and bringing them together.

So it used to be like it is it and ot is ot and now the train shall meet hitachi as an organization brings that capability in terms of merging both these capabilities and competencies together and offering the best out of that.

Now, we do have a very intense focus on r and d and one of the key streams for us in the r and d space has been around industrial and that's very pertinent to the topic uh that we have today on hand. And it's about really how do you apply industrial a i a scale in terms of delivering transformative goals for our customers and helping them really derive the true value from iot whether it be in terms of the cost optimization or in terms of the revenue augmentation or even in getting into new business models entirely altogether.

So hitachi, you know, we started our own foray into what came to be known as iot a couple of years back and out of that, we conceptualized and brought to life. This concept called load and load essentially stands for, it's an acronym which stands for eliminate data, which is essentially shining a spotlight on data and producing rich insights of that data.

So that came into existence as a basic core platform variety from hitachi. But essentially it's a compendium of solutions services products all bundled together. So it's an overarching framework of sorts that we have to the marketplace and does have components like data operations and data governance, data management, cataloging. There's a bunch of things in there.

But what essentially differentiates lama in the current context is all the industrial io solution course that we have really built on top of that. And when pravin alluded that, that's all what we are now kind of repla putting on the a w stack and bring it together part of hitachi's industrial expertise and know how and knowledge along with the cloud platform capabilities of aws to bring to our customers that innovation essentially.

So industry clouds is the next step really in that evolution as we move forward and they are gonna be defining what's the future journey of the cloud platforms

So till the cloud platforms have been largely horizontal and agnostic of the industry verticals, but there's a accentuated need for these cloud platforms to really start offering industry vertical specific functionality. And that's where people like us play where that's where Hitachi really plays into the game. So we do bring a lot of rich competencies across multiple verticals. Be it energy, be it mobility, be it the core manufacturing sector itself and we have 200 plus use cases the solution course that we have developed to address very specific needs across a range of uh uh functionalities like production, maintenance, quality safety. So the the whole plethora of use cases that we have in place and that's the journey we are embarking on with AWS to move that together in that direction.

So logical segue from there, AWS and Hitachi have been partners for long. It's a very strategic partnership that we have in place with AWS. It's a very multifaceted partnership. We are on to about 14 competencies. We have 300 plus certifications, we have 50 plus clients that we jointly engage with together on. I think the core of it is really around the whole industrial space because that's of very key importance both to AWS us in terms of creating the right business impact for our customers.

So we are on this journey together like Pravin said, I'll talk about one of the solutions that gone into the marketplace very recently. But even before that, we are sort of rela a few of our very key strategic core components. So there's asset performance management product that we have in our fall, that's being repla entirely and in dubs to take care, uh uh you know, take use of the core cloud native capabilities that exist and ensure that the whole solution becomes a lot repeatable and scalable and we can offer those to customers at a very logical price point as well.

So this is one of the solutions this is essentially our manufacturing insight solution which is deployed on the AWS platform and uh is also available on the AWS marketplace. And this is really addressing the core needs of the whole manufacturing sec segment at large. It covers the entire gamut of use cases, right from basic descriptive analytics around the forum all the way up to the predictive prescriptive analytics and the complete manufacturing life cycle optimization value chain. And uh it does also offer a rich set of metrics out of the box. So province spoke about oe is an example. We do oe we also do things like on time and full uh defect rates first par ses the health and safety indicators. There's a whole bunch of things that we cover in the gambit of that even production scheduling optimal. So, so it's a complete end to end platform which addresses the need of core manufacturing industrial companies and provides those capabilities out of the box. And this platform has been completely done natively on AWS now. So we are leveraging all the underlying components of AWS in this product. And this is sort of a reference architecture or a blueprint that we have created so that we can achieve a lot of repeatability and scalability in the entire process of doing this with customers.

So before I wrap up and pass it to uh Praveen to invite uh Roy upstage, uh I'd like to take a minute to talk about our collaboration with Penske. So Penske and Hitachi have been very close strategic partners for the past five years, we kind of conceptualized and began this journey together embarked on it uh with a clear intent of identifying real business value use cases that can be impacted by leverage of IoT and artificial intelligence and machine learning. And out of that bone was a particular use case because Penske operates very large fleets, hundreds of thousands of fleets. So maintenance is really a very big cost factor for Penske. And this is what we set out to address. Roy is going to talk more about it. But it's been a very enriching journey for us over the past couple of years. It's not been easy by any means. It has been a journey. That's why I said, and there's no magic wand per se that you bring technology and things happen. It's been arduous. It's been long, but it's been very rewarding and enriching as well as we've gone along there.

So with that, I'll call upon the Pravin back to white rock station. Thank you. I'm super excited to, you know, uh invite Rohit and talk about, you know, what he has been able to do by leveraging some of our services and our partner as you know, computer vision and predictive maintenance, people have been talking about it for the last 10 years, right? But what they have done is truly remarkable and the complexity that they had to deal with and to deliver at the scale that they have done is something you know, um I'm super excited that we're so fortunate to have Rohit here to share the story or to you Rohit.

Alright. Thank you uh Praveen for the introduction and uh Rajesh, thank you for that comment about best for the last. That does not put any pressure on me obviously, right?

So um so show of hands, how many of you know about Penske? Wow, more than I expected, how many of, you know what we do? Ok, great. That's great. So let me, let me introduce uh Penske Transportation Solutions to you, right? So that I represent Penske Transportation Solutions. We are part of the bigger Penske Corp, which is a 37 billion uh organization. Now, most of you may know about Penske because of race cars because of Indy Car. Um but we are part of Penske Transportation, you know, which is again, is, is a privately held organization part of Penske Corporations. Um we were founded in 1969. So right before Covid in 2019, we celebrated 50 years of being in business. Um 3000 locations globally. Majority of these locations are in uh North America and we have about 40,000 associates. This is again, Penske Transportation only what we do, we do full service leasing, we do truck rental, we do contract maintenance and logistics.

So let's talk about that a little bit, right. So how many of you have seen a yellow Penske truck? Ok. So that's our rental fleet. So folks like you and me who wanna move, right? We would rent a Penske consumer, uh, truck, which are those yellow trucks. Um, we also a lot of organizations, large organizations UPS FedEx Walmart during the busy season, which we are in right now where a lot of packages have to move because we all order on the, uh on the internet. So we do commercial rental as well. So those companies actually rent trucks from us. Uh so that's a B to C uh Bb to B model um to deliver, you know, to deliver those goods. So again, in our, in our rental product line, we do both B to B and we do B to C.

Um in our full service leasing uh which is a whole different model. Uh and it is our bread and butter. What we do is uh lease trucks to uh to our customers. Again, an example of this could be where Walmart Lowe's Costco, some of these large organizations, their core business is not transportation, right? Their core business is retail. Uh they want someone else to focus on transportation because that business is becoming very, very complicated. So they come to us, we work with them, we spec the entire uh unit and um you know, the specking is based on a number of factors. It could be the type of payload that they wanna run where they want to run how many miles they wanna run. And then we consult to them, uh, what type of, uh, you know, tractor truck they should, they should buy and then we buy it for them. Right. So we buy it, we place the order with the OEMs and then we do full service leasing as it's called, which means that we lease it to them. But we take care of everything. We take care of licensing, we take care of titling. We take care of uh taxation. If you know anything about trucking, taxation goes down to the county level, right? It's very, very complicated, but we take care of all of that and we take care of maintenance, services, maintenance is again, very complicated business. I'll talk about that, but we take care of all of that. So the, the only thing that the customer that leases from us has to do is put a driver in there and fuel the truck. That's all they have to do. We take care of everything else. That's a, that's a large area for our business and we are leaders uh in this space. And then of course, logistics.

So, you know, most of us are used to drinking Starbucks in the morning. Um some of us uh in the northeast uh also go to Wawa. So the supply chain of all that, most of that is managed by Penske Logistics, which is, which is our logistics organization and then we maintain and manage over 400,000 trucks, tractors and trailers.

So here's an assignment for you. The next time you're on the road, count the number of Penske trucks, yellow trucks that you see and trust me, you'll see a lot more now that I've mentioned it. The only way you would know that it's a lease truck is if you look on the driver's side, there will be a small decal which says Penske. So you get extra credit if you're able to see a Penske lease truck. This was a game we used to play with our younger kids, much better game than "Are we there yet?"

So if you have younger kids, this is a good game to play.

Okay, so like I mentioned, we are leaders in truck maintenance, truck leasing, and this is our maintenance vision and it's pretty simple - it is to increase the customer uptime and to reduce the operating cost. Some of the large organizations that I mentioned, they have their own SLAs with their customers. And they rely on us to make sure that truck keeps on running.

So increasing customer uptime and reducing operating costs are not mutually exclusive, if you think about it. If a truck breaks down, the driver is stranded on the side of the road. We still have to figure out how to get that truck and customer up and running. So we have to either give them a substitute unit or we have to send someone to make sure the truck is repaired. In either case it's a higher operating cost for us. And like I mentioned, as part of our full service lease, we do all the maintenance. So any higher cost hits our bottom line.

So we have to make sure the customer is up and running and our operating costs are low. Now, there is a challenge to this vision and I would say it's a challenge to the entire transportation industry - actually there are two challenges. One is unavailability of skilled labor, in this case technicians, and two is drivers. These are the two biggest challenges that our industry faces today.

Also what happens is a lot of folks don't go to school to be mechanics. We are seeing that the educational side is not being entertained a whole lot by the newer generation. So that's again a problem. The modern trucks are becoming very, very complicated - think about modern trucks as a data center on the road. It's truly a data center which runs on wheels down the road. It's becoming very complicated. It takes a very long time to diagnose a problem and repairing a problem is a specialized skill which is not readily available.

So what I'm gonna talk about today is two use cases where we leverage technology, and these are in production, where we leverage technology and AI/ML to help further our maintenance vision and to kind of alleviate this challenge around finding skilled technician talent.

In 2017-2018, we partnered with Hitachi to talk about this first use case, which is around reducing the diagnostic time and also reducing repeat repairs. I'll talk about repeat repairs - this happened to me personally, I don't know if it has happened to you, but my car had a check engine light. I called the dealership and got an appointment two weeks after I called. I took the car there, they kept it for two days.

After that, they gave me a call and said your car is good, take it out, bring it back in three days because the check engine light goes on. What the technician did was they were not able to figure out the problem so they just did what they needed to do to flip the switch for the check engine light. That's a repeat repair because I'm coming back to the dealership to solve that problem because my problem is not solved.

So our goal, think about this in the context of trucks, if a truck is broken down and the technician only fixes the symptom but not the actual problem, that truck is gonna come back. That's more downtime for the customer, that's higher operating cost for us, which goes completely against our vision.

So our goal with this use case was to reduce the diagnostic and repair time and also reduce repeat repairs. The solution that we built, called Guided Repair, is an AI/ML solution which guides the technician through the repair. I'll double click on this in a bit, but it guides the technician through the repair and tells them what they need to do to solve the issue.

The outcome - we were able to reduce the average repair time by 15 minutes per truck. Think about 400,000 trucks that we manage and maintain, extrapolate this to the scale. Imagine the amount of capacity we've been able to create within our shops because of this technology that we deployed.

Again, like Rajesh said, it was not an easy problem to solve. The other problem we had to deal with was change management, because now you're telling the technicians to listen to the technology, do what the technology tells you to do. So that's a change management problem. The whole cycle was not easy, but very rewarding.

Let's double click on this a little bit. A Penske truck pulls into one of our locations and the technician checks in the truck using our maintenance applications and systems. As soon as the truck is checked in, we know the repair history, diagnostic history, fault code history, we know what type of vehicle it is, make/model/year, and we also know what fault the customer complained about, because that's what started the repair.

What we do is take all that information, transfer the fault code and complaint through API calls to this Guided Repair model. The AI/ML model is already trained with 20 years of maintenance data. It has identified all the patterns and constantly learns - as newer models come in, newer data comes in, it constantly learns.

The Guided Repair model then outputs a prescription, so both Praveen and Rajesh talked about prescriptive - this is a prescription like a doctor, giving out a prescription. The prescription includes what the technician needs to do and the probability of success.

Now the technician can take two routes - one is they can go with the prescription or they can still continue with what they think is the right repair. In both cases, we capture what the technician is doing, and here's why we do that - because if they take the prescription and do exactly what the model says, but the truck comes back for a repeat repair, we want to know that. Because then we need to adjust the model because it's not putting out the right prescription.

The second scenario is the technician completely ignores the prescription and does what they think is right. If the truck still comes back for a repeat repair, in both cases we want to know what action the technician took. So we run analytics to figure all that out, to do training for our technicians and also training for the model. It's a whole feedback loop with this use case.

This went into production in 2019. In 2020 COVID happened, a lot of workers in the US retired early, and we were impacted because of that. This solution really helped us in that scenario. So this has been in place since 2019 and we continue to improve on it.

So this is a reactive use case, it's prescriptive, but reactive because the truck is already at a Penske location, the event already happened. What we started talking about in early 2020 is combining our IoT data - our trucks actually talk to us, sending a lot of data - so we talked about combining IoT data with operational data, repair data, and the goal was to reduce breakdowns. Praveen had that in one of his slides about eliminating or reducing breakdowns.

Because in the first case, the breakdown did happen and it came in for repair. We wanted to reduce the number of breakdowns so the truck is not stranded on the side of the road.

So again, with, through our partnership with Hitachi, we built uh an AI-based uh proactive diagnostics uh model which again combines the IoT data from the truck. This is real time live streaming data with our operational data. And th this, this uh model did go live earlier this year. It is in, in production and we have, and in fact, I'll, this is a quote. I won't read it to you, but I'll tell you exactly what happened here.

So the night that we deployed this um within a few hours, the model identified a truck which, you know, would fail in the next uh x number of miles. So the, the prediction actually is, you know, that the truck may fail in 100 miles, 200 miles, same day prediction, right? So it'll, it'll give uh give out same day prediction. So we were able to, this happened early morning, I think it's like 4:36 a.m. um and, and the model, you know, informed the Penske locations, they called the driver, the truck was pulled into the nearest Penske location, the repair was taken care and the driver was on, on his way in three hours right now, think if we did not have this capability.

So what would have happened is the driver would still keep going. Right. The truck would break stranded on the side of the road, make a call to Penske 24 7, uh, call center. We would then try to diagnose the problem. Send a technician, you know, to, uh, where the, uh, where the driver is, try to figure out the problem. Hopefully, the technician has all the right parts. If not, we tow the truck and you, you get the point, right? I mean, it just takes a very long time, not three hours. So this has definitely reduced, not only the number of breakdowns, but also the amount of time it takes to, to repair a problem.

So let's double click on that, right? So, like I mentioned, our trucks are talking to us uh through IoT. So this IoT data is live streamed into uh what we call is the Penske's data platform. It's a very, it's a very simplistic uh uh you know, image here, but there's a lot going on there because think about it, it's all streaming data. We gotta make sense of it. Um and we gotta, you know, do something with it, we gotta get insights out of it.

Now, this data is then um you know, consumed by the proactive uh model. And again, the proactive model is trained with operational data, like the guided repair model is. So it combines the IoT data and it combines the operational data and then we apply business rules to it, right? So for AI decision making, you need two things. One is the prediction and the other one is judgment because you know, if you just go with the prediction, which is completely dependent on the data and don't apply any context, it may not mean a whole lot.

So I'll give you an example of that. So for example, the AI model may uh predict that a truck may fail in about 1000 miles right now. We know uh that that truck uh is due for service in about two days, right? So it's already there, it's already planned, it's due for service. And on an, on an average, that truck runs 300 miles a day. So 300 miles a day, two days is 600 miles. The truck, the prediction is it will fail in 1000 miles. So do we really need to pull the truck off the road or should we let the driver continue? And then, you know, when the truck comes in for service, we can take care of the problem. So that's the context, right?

So that's the, those are the business rules, those are the smarts, that's our secret sauce that we, that we kind of apply to this and then out comes the uh days and miles to failure. Um and then, you know, based on the, like I said, the judgment, the context that we apply and the miles of failure all of that flows into the Penske's internal and external applications gets communicated to internal stakeholders, gets communicated to customers uh through our digital tools. So all of all of that happens now the brains of this runs on AWS, right?

So the the Penske's data platform, the uh the AI model, the processing all happens on, on AWS. So again, look, I have the privilege and honor of speaking about this innovative work, but the actual work was a lot of collaboration between Penske uh Hitachi and AWS. Uh I mean, those are the real heroes of this. And like I said, this is not an easy problem to solve and we continue to enhance this.

Our goal is to uh continue uh you know, building on this, on this proactive model. Uh because again, right, I mean, uh we have to keep, keep the model uh up and running, it has to learn uh as new as new truck models come out. And like I said, trucks are getting very, very complicated and then we are also working on some other uh use cases um that we, you know, uh in the, in the next year or two, we there, there are some other use cases which are going to again um help our maintenance vision and some of the other um road maps that we have within the, within the organization.

Uh that was my last uh last slide. So, you know, hopefully this was meaningful for you that hopefully that you, you take away something from this. So with that, I'll invite uh Praveen back. Thank you.

Thank you, Rohit. That was an awesome use case. Um so just to kind of uh recap what you have heard so far from Rohit Rajesh and I, and also what is in it for you and why you should consider AWS, right? So there are four things that I want to kind of quickly recap and highlight one is as you know, uh AWS is born out of Amazon and uh you know, we have been doing this for well over 16 years. Um so many of you might know, Amazon is also one of the largest industrial customer. We make products and we have factories and you know, we use all our services that we recommend to our customers as well. So uh so it's something that we leverage ourselves and benefit from. So it's not uh you know, foreign to us.

Uh second thing is we have uh well over 200 purpose 12 services which are serverless. And as you saw here, 14 IoT and also uh four AI/ML services. And I also work seamlessly with other AIML services beyond industrial AI/ML like SageMaker and others, right? And they all are serverless. They work exactly the same way in terms of scalability and you only pay for what you use and you can scale things on demand.

Um so that itself is a huge benefit. And the third thing is, uh we have customers in every industry, every regulated environment, including hospital, military government. So certainly we take security very seriously of your data. So certainly we have industrial grade security and compliance standard that meets all the regions and all the regulatory bodies that are out there on the, on the fourth and important thing is, you know, the scale and the speed at which we operate, right?

AWS is present in close to 250 countries now and we have more than 90 availability zone. What it means is once you pull out an idea and now you're on a scale and scale it and ensure that it also has the speed and the bandwidth. So that's where AWS can really help you in terms of deploying things in seconds, you know, if not minutes really kind of take that benefit. And also if there is any change, change, adoption chain management, you don't have to go through, you know, like the way it takes, once you prove it out, once you make a change, wait for another six months, and that's not the case anymore. It's almost near instantaneous. So it's a huge benefit.

Um you know, as you saw from Penske story and many of our other customer service as well. And I mean, after this, we also have a huge partner network in every geo that we operate. So that's something that you can take advantage of.

Um certainly I invite you to learn more about industrial solutions and IoT offerings by visiting the web page here. And another thing that I would highly recommend is um there is an AWS village we have IoT kiosk, certainly go check it out and then there is an industrial tent at Caesars Forum.

Um so many of our services that you saw, we actually have a demo um that you can kind of touch and feel and kind of get a flavor for how it really works and ask questions and how it is applicable for your particular situation. So I certainly invite you to go check out the Caesars Forum and I think today they close early, so they close at four o'clock. So the last couple of days they were open late. So certainly, you know, there is still time, go check it out and, you know, always feel free to reach out to me, you know, Rohit and Rajesh. If you have any questions, we are here to help you and also listen to your stories. And you know, hopefully next year you can come in and present with us and tell your story like Penske.

Thank you all very much for attending the session today. I also want to thank our speakers, Rohit and Rajesh uh for being so kind and come in and tell our their stories with us. Um you have a wonderful rest of the afternoon. Hope you all had a good Re:Invent, uh, so far and, um, certainly, you know, reach out to us if you have any questions. Thank you.

  • 5
    点赞
  • 8
    收藏
    觉得还不错? 一键收藏
  • 0
    评论

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值