Next-gen trucking: How Iveco Group uses AWS to harness generative AI

text, images, video, and audio is similarly to what human beings do. so today, i'm sharing the stage with my colleagues in particular, i got marco from professional services team and i'm really glad to have a fabrizio and representation of the iveco group.

so let me go through the agenda of what we will present.

so first of all, in, before entering in the solution that we have realized for a product in iveco, we will give a background of the company and we also show the history of this company in order to let you know and enter in the automotive sector.

then we will step back into 2019 when we started to have the partnership between aws and iveco group. and progressively, we will go through the various milestone that we have reached together with professional services.

also, we will deep dive into the process that we have uh applied to our interaction. we will then arrive to the real core of the session which is focused on uh the realization of a new product where general tv i is the real basis. and we will discuss about all the technical details uh of the solution we will finalize then the talk with the next steps that are in front of us and then we are also open for a few questions if any.

so with no hesitation, let me write on straight fabric. thank you alessandro. and uh good morning everyone. i'm so excited to be here. today is uh rain land to 2023. and um i would like to tell you a little bit more.

first of all about myself and my company, i'm fabrics conic. i am responsible for digital advanced technology iveco group and the us in before our last star that as we launched just two weeks ago in barcelona, our renewed uh range of vehicles including electric vans and trucks.

but before telling you more about the veco group, i would like really to start from a question and you see an image here quite clear. so do you know which is the main issue our fleet owners are facing right now? you can think about. i'm sure some of you have already some idea.

so it's really surprising but it is the lack of drivers. so we are really facing a quite um you know, huge problem here in europe. for instance, we have moralized 400,000 unfilled positions here. and uh uh clearly this is a point that we need to address.

so, but what is really the reason why the drivers are more and more uh difficult notifying?

so certainly they are uh away from on long time of their journey. they are facing also unpredictable events. uh we have roads more and more unsafe and also they need to manage uh logistics operations um which are very complex. um and more and more uh under pressure, of course by their companies.

so that's why we need to address these key points and improve the driver experience their life on board, comfort and safety. and today we will introduce to you one of the technology designed to resolve these uh pain points.

so a few words about iveco group, our company is a multinational design test and produce light medium and commercial vehicles, but also power train specialty vehicles for, for application for fighting and military applications as well.

the history dates back to 1975 when iveco was founded, merging several brands including fiat industrial and also magill's. and a few years later, pegas wrote in spain. more recently, we were part of ac nh industry for almost nine years before our spin off in 2022 when we began our journey as an independent company called iveco group. and we had our first day of trade in a million stock exchange, just a few figures.

so we are headquartered in turin. we have eight bands and 35,000 employees worldwide. our 20 industrial sites and 29 r and d centers are spread throughout europe, latin america and asia and the technology digitalization innovation. our our main pillars and we own most of the ip of algorithms and software that we develop and uh as we develop more than 5000 patents, as you can see.

so innovation is the key for us in our company.

now, we all of us have heard about the word perma crisis to refer to a long and permanent period of crisis. i want to propose now a variation of this concept really from a business perspective, when a company faces major disruption in its markets, it can be perceived as basically an opportunity or a threat. and i would like now to talk more about perma chance rather than perma crisis.

i am referring to the revolution in terms of automation that we are going to face. so the automotive sector is of course seeing this automation growing more and more and uh we see now a new interaction between humans and machines that will generate unparalleled business outcomes. and we would like really to tell you more about that today and that automation will generate an improve of sustainability safety. it will make the life of drivers more comfortable as well.

so if eco group has a very customer centric vision since a long time, our company purpose states this clearly. so we are a norm of unique people and brands that power your business and missions to advance a more sustainable society. this is our purpose.

our product and services are around the key terms of sustainability, health safety, and productivity. we believe in the power of uniqueness and distinctiveness as well as innovation and cooper as our partnership with the ws is confirming as well. and our values actually are very aligned with the amazon principles. and of course, this was training our cooperations since a few years.

now. what is important terms of mega trains today? first of all, we are of course working on all of them and you can see in the automotive sector, electrification is the key for sustainability. but also we are opening new business models like paper use which are based on data that we retrieve from uh from the field autonomous driving more and more even if we are talking always about level two plus plus. so with the drivers still responsible of it and a i based services to improve the productivity and the of time. and one of the most important enabler is of course, the cloud, the cloud allow us to collect data, to manage data, to aggregate them and to generate algorithms and services and digital products that help our customers and the is also becoming a marketplace. so in over service transitions.

so our digital innovation journey with the aws started back to 2019 as alexander was saying before when we met in uh seattle and uh we found a common centric vision. so we decided to collaborate for a first innovation workshop we preserve to uh innovate and to generate really a new service starting from the real customer pain points and working backwards to define a possible solution.

in fact, one year and a half later, we launched our driver system that we call driver power at the rain event 2020. and also this approach, as well as the ch pizza team has entered in our company dna. and the way of working was really in our daily experience driving us towards new innovations, but that's not where our story ends.

and alessandro now will describe our aws and professional services support iveco group to implement the cloud strategy platform.

thank you, fabrizio. thank you so much. so when we started the first collaboration, we've seen that the dna of aws proserve and iveco group was pretty, pretty similar. therefore, we try to reiterate the very same prescriptive approach also iteratively for other interaction. and here you can see in this slide, the steps of the process that are the key of the success. and the originally, we always start identifying together with you as a customer, an opportunity, an opportunity for identifying a business problem to be addressed or a problem. and we work backwards from the identification of this problem in order to define which are the business goals that you want to achieve. starting from that we then analyze and propose the collaboration and the different steps in terms of planning. and then we start with what we call start small and think big approach. so we define uh pilot towards mp oc in order to shorten the time to achieve the first concrete result as soon as those results are ready. and we got that point to be presented to the business, to the leadership team, we get a sponsorship to go ahead with the scaling and industrialization. and for this aspect, we uh leverage on the support of the uh customer itself or even our, our skilled partners.

now let me show which are the main achievements that we have achieved so far in the last, just in the last two years. so after uh the first experimentation with ral, we move forward with many other uh streams. the common layer of all of that is the innovation. we have also to keep in mind not only to create technical artifacts, but we have to bring new values and uh support you in order to achieve those business outcomes. so reading from the left to the right, you see that we start from some technical uh achievement that are related to products that have been realized together with the ve a group starting from um web customer portals that are necessary for monitoring the fleet of vehicles as well as monitoring the telemetry and all the other uh necessary information to take the control of the vehicles themselves. and uh uh we launched together the driver pal. driver pal is the vocal assistant companion that allows the driver to reduce the manual interaction with the vehicle itself. so for the vocal assistant, the driver can be more concentrate on the guide. and also there are some nice capability also to get in community with other drivers. but the specific uh driver path is not uh enough to, to manage the fleets. it's important to have data about the fleets. therefore, we concentrate a lot of our efforts also to modernize and migrate uh a connected vehicle platform that allows to collect the telemetry data in a lake. and also to allow the vehicle group to build the business intelligence and machine learning algorithm on top of this data, this is very crucial and marco will cate uh on this aspect. and this is crucial also for how we were enabled on the usage of the narrative a i. then we can say that the group is not only a technology company, they are also starting to produce services and other kind of business model like gate that was mentioned before by fabrizio and is also transforming the ways of working in a more agile fashion to get ready on new trends on the market or new opportunity. today, we will concentrate mainly on the technology and in particular, the ones that i mentioned before.

let's go to see which is the timeline of our achievement. so we concentrate in the last few years. but also we go beyond looking at what i had in front of us, the first basic and most important achievement to start whatever kind of solution or product you have to develop on aws is the foundation. so the combination of all the services related to security and basic ones that are necessary to build on top of it, any business application, this is what we call landing zone, it can be considered as a product because it continuously evolve according to the new product or business opportunity and therefore it will grow, grow again in the next few years.

then as mentioned, driver pal has been launched. the first version has been launched in 2021 and now he is already on the roads. and recently the second version 2.0 has been launched this year, we started to work on the three do with new capability, new feature that will make it more and more uh comfortable for the usage of the driver. as mentioned before we launched in 2022 a new data lake. and this is the basis that allowed driver pile and other tools in iveco group to be uh based on the on this data. but what happened this year, you know, generative i arrive as a wave, there was a risk because potentially it could then completely change our road map. but we did some clever uh decision and taking it as an opportunity to accelerate the development that we were already in place and find new capabilities that could be unlocked

Thanks to the generative AI and thanks to the fact that Iveco Group already owns a huge amount of data where we can base the narrative AI functionalities.

So now the main takeaways that we want to share with you today are:

First of all, giving you in a while, a concrete example of what we realize with the Iveco starting from the business needs that they are clearly in mind.

Second, describe the process that we adopted to achieve this kind of goal.

And third, a reference architecture to give you a kind of kickstarter if you want to apply for similar use cases, the basics AWS services that you might use.

But now to provide in depth details of the first data while we work together with the Iveco Group, let me invite again on stage, Fabrizio. Thank you.

So let's start to tell you more about the driver goals and pain points. So just to be clear, the driver particularly for heavy trucks is staying in the camp long time and he wants to reach destination on time, maximizing efficiency, of course, and reducing the energy consumption.

Then getting reliable recommendation and trustful next steps from the fleet owner and arriving back home to their families and over as soon as possible. So these are the main goals of the drivers.

The pain points for the fleet owner we mentioned before is the shortage of drivers as I mentioned, but for drivers is also the limited amount of real time information in terms of route condition, in terms of weather for eastern condition and to avoid unpredictable events that can impact on their time and generate delays, no, in the delivery as well.

And they need to manage also mandatory stops by regulation. They are also foreseen in Europe. So they need also training and support during their journey including also logistics tasks as well.

So that's why we decide to evolve the driver power through 0.0 to the 3.0 powered by AWS AI. And in particular here, we can see that the solution is leveraging on three new enhancements:

• Proactive vocal prompts, mainly related to real time notification about different scenarios of events. Like you can see also for the icons related to traffic conditions, but also the vehicle health and uses and potentially malfunctions. Thanks to the fact that we develop predictive algorithms able to define the state of health of the main system and components also weeks in advance which allow to notify a potential problem before the breakdown is occurring with any significant improvement on up time for our customers.

• So then a second stream is the local collaborative assistant to enhance driver experience. And this is where the generative AI is really giving a lot of benefits as we can retrieve information for a massive number of documents and manuals to support the driver. No, how to use a certain function in the vehicle, how to make the driving style more effective and safer and many others. And this can be done through a very natural speech.

• And then the third one is the possibility to support multi-language communication. You can imagine that is extremely important as the driver speaks several languages. And, and so the fleet owner can send the message and this message can be automatically translated in the driver language. And this is oon is very useful for all of them.

But now I would like as it was really difficult to bring the truck here on stage, I would like to, to show you how this works. So let's go to Iveco virtually. And then let's see how this system works.

[Conversation with virtual assistant]

So you can you will see more use cases later. But now let us focus really on the benefits.

So this solution is bringing significant business outcomes impacting on productivity, safety of time and sustainability.

In terms of productivity. For instance, this enables smarter routing dynamic ETA improvement, which is the expected time of arrival up to 20% which is of course impacting significantly on the fleet owner business on his direct revenues.

Then we have also an important impact in terms of safety because thanks to the driver pal, the driver can just you know, perform all the main actions, operations vocally, so you can continue stay concentrated on driving and this of course improve significantly the safety and the risk of accidents.

Then we mentioned before the predicted. So we are able to communicate notification on the start of health of the vehicle, but also the driver himself and this can improve the up time and we expect up to 40% reduction of unplanned stops, which also impact on reducing the total cost of ownership of the fleet to.

We can also do any kind of software update over there and this can be notified also through the driver pal 3.0.

When we come to sustainability, which is extremely important for us. It is one of the main pillars. We have also a direct link to the total cost reduction here. The driver pal can help to reduce energy and fuel consumption up to 10% providing driving score updates, insights and training to the driver.

Now let's continue with Marco to deep dive on the technical architectural generative AI.

Thank you. Thanks for being here today. My name is Marco Guerrero and I lead the data artificial intelligence and machine learning and quantum computing teams within AWS professional service.

I wanna start with my personal story. So I, I am a father of three kids. They are crazy about stories. They always want to hear stories and in the past seven years, I use all my imagination. So and now that I'm running out of it, I started to use generative AI to come up with fancy stories that are tailored to their preference, but please don't tell them because they like to think that I'm their hero story maker.

Now generative AI as Alexandra was saying before is a subfield of AI aiming to generate any kind of content across different modalities from text to speech to pictures, videos, music, and the main ingredients of this magic black box are three.

The first one is accessibility to the vast amount of data.

Second ingredient is the computational power that cloud supports and last but not least the introduction of new architectures in the computer science community, namely transformers, which are these giant neural networks that are also known as sorry as light.

Uh there are also known as foundational models. Foundational models are huge neural networks that can perform so many tasks as opposed to traditional machine learning just because they contain a large number of parameters that make them capable of learning complex concept.

Gene T I is revolutionary. But in order to get value out of it, you don't need to train your models from scratch because you can customize, customize pre trained models by leveraging techniques like fine tuning or another very popular approach these days that perhaps in um in these past days, you heard multiple times, this technique is called RAG.

RAG is an acronym that stands for retrieval augmented generation. And later on, I will show you in details how we implemented the RAG approach for driver assistant within the Iveco group, which is the AWS service that can allow you to generate great content and deploy generative AI application at scale.

How many of you have heard about Amazon CodeWhisperer? Wow. Ok. I still see someone who has not heard this this word before, but Amazon CodeWhisperer is a managed service managed service means that you don't need to worry about the underlying infrastructure from networking to security that allows you to design, build and deploy a generative application at scale.

You can easily experiment with Amazon CodeWhisperer and Amazon CodeWhisperer offers you the flexibility and possibility to choose different foundational models coming from the top leading AI companies.

A few days ago, the CEO of Anthropic was on stage with Adam, our CEO to share great news about this ongoing collaboration. But foundational models, even though they have a robust understanding of natural language alone, are not able to complete any kind of business task for that. You still need orchestration of API calls. You still also need to build connectors to access data.

And this point brings me to the next slide which is the importance of having a data foundation, data infrastructure to properly fuel power generative applications.

If you think about it, every company out there, I mean, you you have access to the same foundational models like Iveco Group. But companies that make a value out of it are those who will do so with the proper data strategy.

So moving from generic generative AI to generative AI that are powered by your data is key to your success. And with the Iveco Group over the past few years, we built what we call AWS connected mobility, the lake that supports the full life cycle of the data from ingesting and collecting real time or near real time data coming from the vehicles all the way to querying and analyzing data using analytic services or machine learning services.

As you know, Amazon SageMaker is our workhorse for machine learning.

Last point, I would like to stress on this slide and I'm very sensitive to this topic. Is responsible AI? And I was very happy when I started to work with Iveco Group and especially Fabrizio, when we we started to discuss about how to build responsible um AI within Iveco Group ecosystem.

So responsible AI is a central, is, is, is an imperative strategy to AWS. And when you decide generative applications, keep in mind concepts like trust, safety and transparency. And this is what we do all the the every day basically with the Iveco Group aligning the application to Iveco Group values.

Now, after this few slides, let's switch gears and dive deep into the functional architecture of Driver Pal 3.0 which is this next generation of conceptually, we can structure the solution with five main building blocks starting from the left all the way to the right.

Let me start with extremes blocks which are the client application that is the main interface through which a driver communicates with the the the application. It is installed in the infotainment system within the truck cabin.

If we now jump to the device, a connectivity layer, we can appreciate the fact that this layer handles on board computers and sensors that retrieve and ingest data coming from the vehicle. It also responsible for the connectivity aspect. For example, in order to be on the cloud, you need to access internet through 5G modems and so on.

So while these two layers, these two extreme layers play a crucial role for the entire application. Today, I would like to focus on the three central layers, the API layers, the generative AI layer and the data layer. Here is the remaining transcript formatted for better readability:

And in the next slide, I will walk you through a target architecture. Let me first tell you that this is a simplified version of the high level architecture that we design in collaboration with the Iveco Group. So we always co-create with the customer.

So again, we start with the API layer and then we move to data layer and last but not least the intelligent core of the application, which is the degenerative AI layer.

The API layer exposes all the capabilities to query data, invoke predictions, integrate with external systems and so on. We rely on REST APIs and event driven APIs.

Then the data layer leverages AWS database and analytics services to ingest, store, process and query data. As I mentioned before, the data lake plays a crucial role here.

And then we have the AI layer, the magic, the secret sauce. This layer leverages Amazon CodeWhisperer to power the most advanced natural language capabilities. So let me tell you a bit more about this layer.

We decided to adopt a hybrid approach. We have some use cases powered by a classic retrieve and rank machine learning approach. So you query a knowledge base.

But most advanced use cases like providing driving assistance and answering complex questions are powered by RAG, as I mentioned before. RAG stands for Retrieve, Augment and Generate.

So let me tell you how it works in practice. The driver asks a question to the assistant through voice. The question is transcribed with Amazon Transcribe.

Then we use Amazon Kendra to query relevant passages from manuals and documentation. Kendra returns context passages relevant to the driver's question.

These passages are then used to augment the driver's original question. So we have now an enriched text.

And this enriched text is finally fed into CodeWhisperer that generates the most appropriate response.

As you can see, we are mixing the power of search and natural language generation. And this allows us to provide really tailored information to the specific driver's question leveraging all the documentation available.

The last point I want to make on this slide is that we continuously improve the solution leveraging usage data. So we collect anonymized conversations and we use them to fine tune the models and make the assistant smarter and smarter day after day.

So in summary, in this presentation we described the end to end process we followed to build Driver Pal 3.0 for Iveco group leveraging AWS AI services. We discussed the business needs, the architecture and the benefits this solution is bringing.

I hope you enjoyed this session and got useful insights to bring AI in your products. Thank you!

So as you can imagine, there are a lot of services that come together to realize the solution. So API layer you see here a bunch of services. Let let me walk you through each of them.

We start with Amazon Route 53 which is this highly scalable service for DNS (DNS stands for Domain Name System translation).

Then we have the security service AWS WAF. It is not a waffle. OK. WAF stands for uh web application firewall and it protects the API Gateway from web exploits like SQL injections or cross site scripting.

Then we have the Amazon API Gateway that is, is responsible for API management and it represents the front door to the Gen T I layer.

And last but not least we use Amazon Cognito for user authentication um and authorization.

We move them to the data layer. So we have a bunch of services here. So let me again walk you through each of them.

We store telemetry data uh in Amazon Redshift which is the AWS service for data warehouse. It is a, it is a, it is a, it is a structure um big database that also stores the output of machine learning algorithms that Iveco Group has designed and built over the past 20 years, those algorithms run against the raw data and they output something like predictive maintenance recommendations.

So the output of this algorithm are stored in Amazon Redshift and later on, it will be clear why we store also this this output.

Then uh once we store this uh this uh uh telemetry data in the output of this machine learning alive. in, in uh in Redshift, we move them to DynamoDB which uh uh but before that, we use AWS Glue to run ETL jobs. ETL stands for Extract Transformer Load to further process this data.

While doing that, we use the Iveco Group data catalog for meta data enrichment. Now let, let me give you the gist behind this.

Um some of the data in the store in Amazon Redshift are not human readable the tax, they don't have any business context attached to them. By leveraging the business grocery in the Iveco Group data catalog, we are able to make like uh not human readable tags more business friendly.

And this is extremely important to further inform Amazon Kendra that is still smart. But if we can help the smart people with even more intelligence, why not? And this is the key.

Finally in Amazon S3, we store and structure data like vehicle manuals and uh anything in in pdf format that can be useful to retrieve contextually relevant information.

Now, this is the final layer which is the most important one is called Gen TAI layer. It, it consists of different services.

We start with the the multimodal agent that's routes the queries requests coming from uh the user from the driver. And uh it used Amazon Kendra for indexing and retrieving contextually relevant information from vehicle manuals. It also triggers queries to Amazon DB to retrieve structure relevant data.

And in the next slide that this this concept would become super clear, i promise uh last but not least we have Amazon SQS that is responsible to decouple the conversation of workflow that happens above from the permanent storage of the multi term conversations.

But before starting the historical conversations in Amazon History, we also use Amazon Comprehend to detect and reduct all the PII data PII stands for Personally Identifiable Information. In this way, we are compliant with the Iveco Group regulations with the GDPR. And as i was saying before, in this way, we embrace the responsible AI principles.

Now, the next slide will walk you through a concrete scenario. So we will try to um help you um uh walk you through the, the real experience of the solution.

So for this example, we are, we are using a RAG approach, RAG approach grounds large language models to verifiable information coming from the Iveco Group data lake and also offers some insights on the generative process behind the large language models.

So let's suppose that the driver will ask the question that you can read on the left side of the slide. "What is the ECO switch and how can it help me?"

Driver Pal power by by Amazon Kendra will start to reply as you can see the answer on the right side of the slide.

So i wanna really now focus on what makes this application reason over the question that the driver asked as you can see from the answer Driver Pal was able to first of all identify the definition of the ECO switch within the user manual.

Second was also able to retrieve all the historical information about the usage of EGOS which from the structure telemetry data stored in Amazon DB.

So that is why here we we we we talk about multimodal agent. We have two modalities. One, the first mode is the first modality is unstructured data, vehicle manuals. The second modality is structured data, telemetry data, real time data coming um from directly from the vehicle and stored in Amazon DB.

So last bit on this light is what really makes this application a reason like we the human being do is of course the great foundational models that are provided within Amazon Kendra but also a a very powerful technique for prompt engineering, which is called chain of thoughts chain of thoughts.

If you think about it, it's um is an outcome based instructions that can elicit the response that we want from a large language model. I will be happy also to address any technical question after this talk.

And uh my last slide before moving to Italy virtually again is about security and privacy at AWS. We say that security is job zero. Now with generative AI, we can say that responsible AI and privacy is job zero.

So uh within this application, we make sure that no Iveco Group data are leaked from this application and are used to train um new models or even existing models within Amazon Kendra. All the data is encrypted address and on transit. And the again, i i it was a pleasure to really embrace the security, all the security aspect behind this application with the Iveco Group and their great engineering team.

Uh with that again, let's watch the second and last part of the video and let's fly to Italy. Of course, virtually.

How is my driving style based on the provided context? Your driving style seems mostly appropriate with a few areas that could use some improvement. Specifically cruise control usage could be improved by exploiting it more often whenever possible. The accelerator pedal usage needs work. Try to reduce harsh acceleration occurrences and press the pedal more gradually. Usage of the ECO switch and ECO roll features should be improved by keeping them enabled at all times during driving Overall, focusing more on using cruise control, driving more gently and enabling eco features should help you optimizing your driving style.

Hello, Kevin, let me give you some information about your daily trip. Your vehicle's health is good. The weather along your route will be perfect with 24 °C and clear skies. Nowadays, your fidelity program level is expert. How can i support you today?

Can you tell me out and from my level? Sure, you can work on many different aspects to improve your level from reducing the number of harsh braking to increasing the usage of predictive cruise control. I can provide you all the details if you want as an example, keeping the ECO roll active during your daily missions will grant you the master level in about two weeks.

Great. Let's do so. Due to a car accident in 70 kilometers, the delivery plan has been updated. Your new destination is Milan MP X two. New estimated time of arrival is 230 pm, one hour, 15 4 mil delivery warehouse is already informed and waiting for you. Have a safe trip.

Ok, let's navigate to the new destination. Thanks for optimizing my time. My pleasure, Brenda, always a lot to learn from uh Marco.

So thank you very much. It's always a pleasure to hear about architecture and Gen T you have seen that before. Also talking about fidelity program, i would like just to spend the one word about that.

So basically, we define a way to provide prices to the best drivers in terms of driving style, which is also a way to have the attention of drivers and also win to win with the fleet owners as they can improve full consumption, energy consumption and win prices because they are doing well.

Um so this is what about the fidelity program we have seen before? Now, um i would like um really to show to share with you uh a quote of our co gerri marx uh which is underlining definitely our stronger partnership with the AWS, which is focused on innovation and the distinctiveness because we focus really on the, on disrupting the market with new solutions.

And he is really obsessed also about our customer centricity as he decided as well to have the iconic amazonian empty chair in his meeting room. You always represent our customer, but not only external customers, but also internal customers, internal shareholders.

Uh because all the company must be engaged in our digital transformation. And uh uh just as a confirmation of what i was saying, we are expanding the application of generative AI uh from our products, but also to many others uh departments and processes in our company.

So we have additional use cases uh to make uh our process more efficient. For instance, we can create a chatbot uh to improve the documentation and knowledge management in engineering. This is helping a lot.

Also, we can allow um an intelligent automatic uh tender analysis which is also important to support our business units. Uh also making possible automatic coding to help uh our software developers and apply computer vision to enhance uh smart manufacturing and many others that we are working on.

But now let's come back to LA for the conclusions.

Thanks for this. Thank you Fabrizio and you have seen how many things are really in front of us. So it's really just the beginning and uh we are really happy about, we started to adopt generative AI this year, we need to continue this kind of partnership we preserve with Iveco group, with partners to scale this solution and to continue innovating with Iveco group.

So this is a journey where proserve are very crucial. But i would like just to mention also uh the huge amount of initiative that the AWS is launching this year, not only on services but also on programs.

For example. Uh recently, the AWS Generative Innovation Center has been launched composed by a spectrum of expert in uh machine learning and specifically on generative AI that can come to help you to develop uh your, your business application together or uh or not with the Professional Services team.

So now we are arrived at the end of the session, i would like to invite on the stage uh Fabrice and Marco here. I would like to share our uh our links. We are happy to uh get in contact with you uh today in the next days in the next weeks, we got your uh uh also information related to the connection of professional services team.

And beforehand, i remind uh to compile the survey on the application. And let me thank you all for having attended the.