AI2FUTURE: How Data Could Save the World
Zagreb is one of the emerging innovation centers in Europe. The city is home to the country’s largest university and has a strong technology and developer scene. Therefore, it’s ideal for driving innovation across borders. That’s why one year ago, Porsche Digital opened a new office in Zagreb as a joint venture with the Croatian tech company Infinum. To further network in this up and coming tech region, Porsche Ventures and APX are partners of this year’s AI2Future conference.
The two-day event welcomed the best and most visionary AI experts, about the importance of data science and dialogue for their work. Two Porsche digital colleagues, Mirna Marković, data scientist, and Ema Šimon, Product Owner of the data science team, were presenting their work as keynote speakers. We asked them, to tell us more about their vision of how data could save the world.
“Going back to basic” is the title of the talk we gave at AI2Future in Zagreb. At first glance, the terms “going back” and “future” do not seem to go together. But for us, a good future, especially in the AI sector, requires a reflection on the essentials, namely the collaboration of data science teams and engineers, as well as to focus on really understanding and using the data we collect. The guiding principle of AI2Future is: Join the revolution!
As Porsche Digital, we must participate in this AI revolution in order to be able to shape it. We are convinced that we can only successfully take this step forward if we first take a step back and recognize how we are connected — both on the technical and on the human level. Reflecting on these connections of science and practice, of engineers and data science teams, provides us, in our opinion, with the crucial prerequisite for more quality. Improvement in data collection, analysis and data quality can only be achieved if we work well together. This applies to Porsche worldwide and is also evident here in our networking, for example with our Infinum joint venture at our site in Zagreb, as well as in the microcosm between data scientists and engineers.
The value of data-driven strategies
In our contribution, we put forward the thesis that data (can) save the world. By really focusing on data and how we collect it and process it, we also overcome the current challenges such as lack of data connectivity and poor data quality.
And if we may not save the world with data science in the first step, we will at least save the car industry: because data science can redefine the automotive industry by focusing on customers’ real needs and desires. Due to the new validated data, vehicle development becomes both simpler and more custom-fit, and commerce can be controlled in a more target-group-oriented manner and is thus facilitated. To ensure efficient execution of these opportunities and that all the necessary technical requirements are met, we as product owners and data scientists advocate three pillars of collaboration: cooperation, expertise, and experience.
Cooperation: Data science teams cannot work meaningfully on vehicle-related projects without extensive and in-depth cooperation with colleagues from automotive engineering.
Expertise: It is crucial to define what data is needed, how to interpret the results and which outcomes are preferred — again, collaboration with the experts is one of the basic requirements for success.
Experience and time: Over time and with increasing experience, the data science teams gain a deeper knowledge of the field and are better able to support and collaborate with the technical teams.
In addition to these three pillars, it is data science itself that helps a data-driven project succeed with its unique perspective. For example, data scientists support the engineering teams by asking the right questions and maintaining dialogue between the different functions. This creates an iterative process in which the different teams work together to find helpful solutions. This process is also important because, especially with static car components, you can’t keep readjusting, they are statically fixed, cannot be flexibly converted, but have to take the right path from the start.
The challenges of data science in the automotive world
In this process, data scientists are confronted with the challenge of two worlds: Technicians and engineers are used to the speed of the car market and have a hard time with the way agile software developers and data scientists work. Similarly, data-driven engineering faces the challenge of wanting to support the vehicle development process in agile mode while being highly dependent on traditional vehicle development practices. This leads to a work process that starts in a very rudimentary way with the definition of data collection, exactly as we requested at the beginning of the text.
80 percent of our work energy and time is therefore invested in the following three tasks:
Understanding the business model: We develop possible use cases for the application of data science by evaluating both the technical possibilities and the business benefits.
Data understanding: We conduct preliminary data analysis and exploration, looking at both the statistical properties of the data and the data quality. This is not a traditional data understanding setup, as we are also participating in the process of data collection, as well as the preparation, analysis and exploration of the data. All things considered, this is an iterative process and the only way we can bring quality data.
Data preparation: We select the features to be used for modelling and analysis, clean the data and integrate and aggregate different data sources.
These three preparatory steps determine the success of data science projects. This is because the quality of the data is of far greater benefit to the overall project than the quantity. High-quality data is at the heart of any machine learning project and collecting high-quality data from vehicles is complex and time-consuming. That’s why it’s insanely important for data science teams working with vehicle data to be involved in data definition and quality assurance. At the same time, cooperation with technicians is also important for data evaluation, because they have the know-how to evaluate the complexity of the data.
InnoDrive: a great example of collaboration
One of the projects where you can see that good cooperation between the different approaches leads to an outstanding product is the Porsche InnoDrive driving assistant, an assistance system that optimises speed and anticipates ahead, making it the essence of intelligent driving. Two data science teams were involved in the development. While the first focused on developing the product, the second focused on validation. As the new team was not biased in its analysis of the current product, additional insights into the product could be gained. Data collection was a lengthy and iterative process that took more than a year and required a lot of data quality testing and expertise.
The final data set included only 16 different attributes, from more than 100 000 possibilities but it provided many insights showing that data quality is important for delivering value. The final product of the analysis was used by many different teams and stakeholders who were connected through the work of the data science team.
To conclude:
There is no progress without collaboration, and we cannot work without each other if we are to truly champion data-driven business models. Vehicle data is not easy to collect. Therefore, we should focus on quality and not quantity. Solving a business problem is a key to working together successfully in terms of quality.
Many large companies offer complex and exciting projects that can only be solved by looking at the connections.
Data understanding and data engineering are key components of data science. Therefore, these teams should support and be actively involved in data collection. Because over time, they can bring additional benefits to traditional vehicle development — from new perspectives to the dissemination of knowledge across different teams and organisations. Because sharing knowledge, data and resources is the only way we can all move forward — until the world is saved.
About this publication: Where innovation meets tradition. There’s more to Porsche than sports cars — we’re tackling new challenges, develop digital products, and think digital with a focus on the customer. On our Medium blog, we tell these stories. It’s about our #nextvisions, smart technologies, and the people that drive our digital journey. If you want to know more, follow us on Twitter, Instagram and LinkedIn.
