In this episode, Martin Gutberlet, Industry Principal EMEA for Manufacturing at Snowflake is joined by Thierry Martin, Head of Enterprise Data and Analytics at Toyota Motors Europe. They discuss Toyota's evolution towards software-centric operations and the implementation of data mesh architecture, and Thierry highlights the company's culture of continuous improvement through AI upskilling programs and hackathons, and shares about the pivotal role of a robust data marketplace in optimizing manufacturing and supply chain operations.
In this episode, Martin Gutberlet, Industry Principal EMEA for Manufacturing at Snowflake is joined by Thierry Martin, Head of Enterprise Data and Analytics at Toyota Motors Europe. They discuss Toyota's evolution towards software-centric operations and the implementation of data mesh architecture, and Thierry highlights the company's culture of continuous improvement through AI upskilling programs and hackathons, and shares about the pivotal role of a robust data marketplace in optimizing manufacturing and supply chain operations.
---
Calling all developers, business leaders, IT execs, and data scientists! Snowflake World Tour is your chance to learn and network. Discover how Snowflake’s AI Data Cloud can transform your career and company. Experience the future – join us on tour! Learn more here.
[00:00:00] Producer: Hello and welcome to the Data Cloud Podcast. Today's episode features an interview with Thierry Martin, Head of Enterprise Data and Analytics at Toyota Motors Europe, hosted by Martin Gutberlet, Industry Principal EMEA for Manufacturing at Snowflake. In this episode, they discussed Toyota's evolution towards software centric operations and the implementation of data mesh architecture. And Thierry highlights the company's culture of continuous improvement through AI upskilling programs and hackathons and shares about the pivotal role of a robust data marketplace in optimizing manufacturing and supply chain operations. So please enjoy this interview between Thierry Martin and your host Martin Gutberlet.
[00:00:45] Martin Gutberlet: Welcome to the Data Cloud Podcast, where we explore how leading companies are transforming their business. Today, I'm thrilled to welcome Thierry Martin, Head of Enterprise Data and Analytics at Toyota Motors Europe, who brings a wealth of knowledge from pioneering data and AI initiatives at the world's largest automaker.
Thierry, welcome to the show. So, you have been at Toyota Motors for more than 23 years, an incredible tenure that spans massive industry changes. Today, we are seeing OEMs shift from hardware first to software centric operations. Can you explain more about your team's mission and vision and how you're driving this transformation at Toyota?
[00:01:36] Thierry Martin: So my team's mission is actually, and the mission is to enable anyone in the company with valuable data products and analytics. So we are a very big OEM, so we have 8 manufacturing facilities in Europe, so we are selling more than 1. 2 million cars per year. So there is a massive amount of data coming for looking at the parts and then what, uh, delivering vehicles, the logistics subs.
So there is really a massive amount of data, but so far we have had difficulties in gathering all the data together. to have the right insight from the data. So we have a lot of legacy systems, actually, and then now we realize that if you want to enable AI and transformation, then we need to have actually access to the data.
So that's, that's one of my team priorities is to establish the data mesh. Actually, that's what we are building on, on Snowflake. A second part of my second branch in my team is taking care of the analytics, so bringing machine learning models into production. And then for that, the transformation part is also to talk directly to the business and to help them to actually build machine learning pipelines.
So we have also new roles in the team, like data translators, and these people are there to, to make the bridge between. between the business with business question and with data scientists, we would speak a totally different language. So we have even some new rules in the organization to, to make sure that we have the right, the right communication.
So that's, that's a part of this team. Another part is also communication because we have to communicate a lot about data. Data and about AI, because everything is very new. That's also something that we have to do in the company, in the transformation, is to communicate a lot about, about AI. And my third team is DataOps, MLOps.
This is also something new that you had to create in the, in the last year to build a team that can deliver machine learning pipelines.
[00:03:34] Martin Gutberlet: Very good. And change and communication definitely belongs to each other. Thank you. Toyota's DNA is rooted in Kaizen. And the Toyota production system. Can you explain this philosophy?
So how have you translated these foundational principles into your data strategy? And how does this strategy translate into how your team builds data products and data mesh architecture?
[00:04:02] Thierry Martin: These principles are really part of our company culture, and this is something which is very, very strong at Toyota.
So actually, when we, when we hire people in the organization, the first thing they receive is a training about TPS. And even we send them to the production line, the factory, one of the factories, to actually build cars, to understand what is, what is the end and what is the line that we can take. To, to stop the process, to not deliver bad quality to the next step.
So all these principles of Kaizen continuous improvement, the, the way we solve problems. So we have the Toyota Business Practice, a very structured approach and standardized on how to solve problems. It's really deeply ingrained into our company values, and that's really what we can see our company culture.
So even working in data, we try to implement that as much as possible. Toyota is very famous for the quality, so we deliver. Vehicles with excellent quality, but the same principle we should apply also to data. So of course we can, we can create a data pipeline, but if, if the data is wrong, or if the data is old or not refreshed on time and things like that, then we deliver also bad quality.
And that's part of the, this principle of TPS. We should not deliver bad quality to the next stage. So that's why it's very important to have a very clear image of what we have as products in the organization. And for that, we developed a full set of process, actually, to create data products and to deliver quality data products.
Very important also for us is the, is the visualization of what we do. So the, the process are visible. We also created, for instance, a, what we call, Call it a tube map. It's like a metro map. It's just a visual, like if you look at the, uh, the metro map of a city, you will see different tracks, which are connected down with different stations.
We do exactly the same for the creation of data product. And then we have different lines, one for the, if you want to create an AI model or a BI or just a data product. So these lines intersect and on these lines, we have different stations leading to governance and, uh, some approval to the release in the, in the So that, that's part of this process that we want to standardize the process as much as possible and to make it simple and visual.
[00:06:17] Martin Gutberlet: Great. Thanks. To apply Kaizen for data quality. This is awesome. Thanks for sharing. You're producing like Toyota produces more than 10 million cars globally every year, right? So this is massive, right? So how does your data strategy support operations and transformation at this massive scale? And do you use any industry framework to support this level of production?
[00:06:42] Thierry Martin: I will talk for Europe because that's, that's what I know most. So we produce 800, 000 cars a year in Europe. So we sell 1. 2 million. Yeah, of course, the data is used a lot in our supply chain and in the logistics area in manufacturing. I can give you some, Very simple example of what we are doing with it.
For instance, we are building predictive model to know which accessories will be fitted in the, in the hubs, for instance, to predict the workload, or we can have also some, uh, some models to help to optimize the loading of trucks from our distribution centers, but these are quite. Classical AI example, not transformative, but what we are looking for right now is efficiency improvement.
So it's really the spirit of the guys and small improvement, but we can find from, from data. The next step will be the more transformational approach, but right now we are in the, in the small step by step improvement. In manufacturing, we are also starting with the IOT data. Of course, there we have a massive amount of data, so each factory has as many robots for the welding and this generates a massive amount of data, which is streaming.
What we are doing is trying to also look at what we can do with this data and we have Practical examples like using the data from, from the factory, how we can optimize or reduce the energy consumption in the paint shop, which is the number one consumer of energy in the, in the plant. We can see huge benefits by using, by using data.
Still, we are at the beginning of our journey. Yeah. So we are now the phase where we are building many data products, and especially for manufacturing, if I'm not wrong, we have about a hundred and data products in the pipeline so that we, that we are building, uh, and this data can be maintenance data, quality data, production line data, but also IOT data. We are building all that right now, and then now we are start, we are starting to make connections and getting value from the data. Yeah.
[00:08:47] Martin Gutberlet: Toyota has also, you know, been fostering a culture of innovation through initiatives like you do hackathons and you do AI upskilling programs. So what impact have these initiatives had on employees innovation and skill development?
[00:09:01] Thierry Martin: That's very important for us. So people development is really at the core of Toyota values. And even talking about AI, we don't see AI replacing people. We see more AI empowering people. augmenting people. So that, that's, this is the concept that we are, that we are following. In terms of upskilling, we have several training programs and we have some coding classes that we can find online in the, in the company.
So this is more if you want to transition people from a current role of an engineer or analyst towards a data analyst or data scientist. So we have this track possible in the company. And then we have Also, AI workshops that we organize, and these AI workshops are really hands on workshops. It's about four to five hours, and then we explain to a group of people, typically in the setting you would have like 30 people, and we select one division.
Of course, they volunteer to come, and then we provide an actual hands on class on what is data, what is AI, and see practically, okay, this is a table coming from that we uploaded from Excel or from our dataset. And then we teach people how to build a machine learning model from end to end, and we even deploying it.
For that, we use our machine learning platform for that. And we can see a very, very positive response from the people. So because a lot of people say, I understand what is data now, I understand what is AI, and we understand also, it's not that difficult. So we try to remove all the jargon of the data science to make it really easy to understand.
Approachable and understandable by the people. And we'll not only give this training to young graduates or, or employees, but also to execs, so we have VP level and also EVP level who also follow the training and they say, no, I understand this, I can do that. even my members can do that. So we go that far.
So this year we trained a bit more than 200 people on that. And we can see that after the training, people request to have the credentials to access our data mesh. And then they start also to come with ideas that they raise. And then, then we start discussing with them about the value of the idea and if you should bring it or not to production, do we need to prototype it or not?
And we try to really do have this citizen development approach, where people learn new skills and then they, they try directly this data, which is available right now. For the third one is the, the hackathon. So we also organize a yearly hackathon event. And so we do that each year in a different city in Europe.
And typically we are going to bring together about 60, 70 people from different organizations of Toyota. So we will have people from, from the manufacturing plant, from logistics area, from R& D, but also from markets like Toyota Germany, for instance, or Toyota Italy. And then these people come and they come with, um, Uh, it's more like a bottom up approach where they come with ideas and then for three days then do, they do data collection, data governance also is part of the, of the activity of the hackathon and build also a data analysis, basically.
The idea is to come with a pitch that is then presented, presented to, uh, to execs. And actually this year we invited, actually, Snowflake to the event and to support the, uh, the activity, uh.
[00:12:38] Martin Gutberlet: Excellent. That's awesome. And what I really like is that you have said, you know, AI is not replacing, AI is actually empowering people to do better. And that you do training across all levels, including boards in your companies. So really, really cool stuff. You're also creating a data marketplace for internal teams to browse and request access to data, right? And how does this marketplace model work and what are the key benefits for business users within Toyota Europe? And how do you think about governance within this project?
[00:13:10] Thierry Martin: So these are good questions, huh? So I will tell you a bit about my background first, because I have not always been in IT. Actually, I come from R& D, and for 20 years, I've been actually developing cars. But I did a move towards data science and data engineering, and then now I moved to IT.
But actually, my first job in IT was to start some data science activities. But then my first question was, where is the data? And okay, there is no data marketplace, there is no data cloud, there is data a little bit scattered everywhere. And that was really the biggest problem for me is, where is the data?
And you know, the data is available because it is in mainframe, it is in AS400, it is in MSSQL. So a little bit everywhere, but actually there is no feasibility on where is the data. So one of our priorities is to build a data marketplace. And that data marketplace is actually on top of the data catalog, so it's, it's all part of the same solution.
And then we have a full governance framework in place to publish data product into the data catalog. And then from the data catalog, we promote things into the data marketplace. So when you visit the data marketplace, you can then see, okay, what's hot, what's new. I want data for, uh, for quality, for instance, then you can, you can browse and arrive to the quality area.
And when you start browsing for your data products, then you can see the full detail because it's part of the data catalog. You can see the lineage of the data. You can see who's the owner, what's the definition, what is the source of the data, what's the quality. There are lots of ratings like this. So there is really a wealth of information.
All the metadata is actually there. From there, so that will really help the people, the consumer, to easily get access to data. After that, how to get the data, that's also part of the data marketplace. So you actually shop for data, it's really like a web shop experience. So you go there, you say, okay, I want to request, I want to put this data product in my basket.
And then when you submit, you check out your basket, then it goes to the data owner, who will receive in part of the workflow, your request to access the data. And if there is some PII data, like privacy related information, personally identifiable information, then there will be also a trigger which will be sent to the legal representative for the data protection manager to see, okay, there is a request to access this data product.
It has been granted, but okay, we need to check the consent. And then all these people have to give their consent. their approval so that you can get access to the data. And when all the access is approved, then finally there is somebody in the backend that will then link your user ID to the right Azure ID group that will be then allowed to access your data in Snowflake.
[00:15:59] Martin Gutberlet: You know, with so many manufacturing facilities in Europe, And you already mentioned some around 180 use cases you have. How are you using AI to optimize production and supply chain operations these days? And what are some of these concrete examples you'd like to share with us? And, you know, maybe the last one, you know, do you also do something around ESG within these initiatives?
[00:16:21] Thierry Martin: For the first one, I already mentioned some, some use case that we are, we are doing, uh, like the, the truck loading optimization. So this is, which is very useful actually for the people who have to, to load trucks and they see the, all these flows of parts and how to optimize the truck. So we are using optimization algorithms for that.
The optimization of the workload in the hub for the people, because you have to plan in advance. And actually, you know, all the, what, what's the volume of car coming to the hub, for instance, one month in advance, but how to plan the accessory fitments. This is the kind of things that we can do with ML, which is quite easy.
quite good accuracy. Yeah, in the manufacturing. So far, one of the biggest ones is the energy optimization, but actually use cases are starting really to pile up. Actually, if I look at the use case, we have more than a hundred use cases actually in the pipeline right now. Some are data analytics rated, but more and more we have also GNI related like chatbots, yeah, access to knowledge, to unstructured data.
We see more and more of this kind of use case coming. And now we are. We are sorting out which ones are the priority ones. Regarding CSRD, of course, we are also planning to use the platform for that, for the reporting, because that's now becoming a mandatory obligation to report.
[00:17:43] Martin Gutberlet: Good, good. You know, ESG reporting will get more intense over the next couple of years. So that means it's not only going to be on plant level, it's going to be on product level, on even single assembly level, so therefore it's good to have data at a single place. Looking ahead, so how do you see AI evolving in the automotive industry and what kind of emerging opportunities or challenges do you see as a leader?
[00:18:12] Thierry Martin: Well, I see that AI will be more and more prevalent in the industry and also in the, in the vehicle. The challenge I see is really to get the data, which is, before it was data ready for, for BI, for BI to have the right, the right data so that we can trust data, which is already a very big thing. And we are not.
They're yet a hundred percent, but now we have AI, GenAI, also have to collect the data, but also the metadata about the data so that it can be used by GenAI. So how to make the data GenAI ready? I think that will be a challenge and it will be, it will be solved by a proper governance, I think. So having the right pipeline, but also having a perfect description of, of the.
of the data and what it is, what it can be used for, what it cannot be used for. And then with this, then we can start to have, we can start to trust also some Gen AI models to, to work with it. So that's one technical challenge. The second one is, I think we can go back on the, on the people is, is, is keeping the, uh, pace with the, with the technology.
And so we need to continuously develop. People, we need to have the right tools, which are evolving well, but we have to keep people up to date with using these tools and with the new technologies. So that's, that's very important. And also how to retain people. That's also a big challenge. Yeah. And yeah, it's a suggestion for any data leader, I would say, get your hands dirty or so on.
So do coding by yourself to understand also the platform. So personally, I'm also coding in Python. And I'm sometimes making a streamlet in Snowflake to see how it works and can I get access to data. I'm testing actually myself, the platform, using the eyes of a data scientist or data analyst or business customer, just to see that everything works well, because being customer centric is actually our, yeah, that's also part of our values.
[00:20:06] Martin Gutberlet: Fantastic. So it very much means get your data right and then you're getting your AI right. Yeah. And while doing so, keep hands on. Thanks, Thierry. Yeah. Really, many thanks. Yeah. Thank you very much.
[00:20:19] Producer: Calling all developers, business leaders, IT execs, and data scientists. Snowflake World Tour is your chance to learn and network. Discover how Snowflake's AI Data Cloud can transform your career and company. Experience the future. Join us on tour. Learn more at snowflake. com slash world dash tour.