BMW Group innovates with Azure for 10 times more efficient data delivery

Luxury BMW automobiles are designed to deliver high-performance and convenient mobility that’s both enjoyable and practical. Customers trust in the reliability and quality of the cars’ complex systems and incorporated features—from voice-activated assistants to 5G connectivity—that keep drivers safe and confident on the road.

“BMW is a master of integration,” says Sebastian Heinz, Mobile Data Recorder Co-Creator at BMW Group. “Our focus is to create a car that really fits the customer. That requires integration of thousands of digital components.”

Operational testing of those functionalities is performed using a fleet of 3,500 development cars. Before 2018, operational data from the cars was captured using onboard hard drives manually transferred to on-premises servers. Engineers had to wait at least a day to get fresh data for analysis. That slowed design and prototype times in an industry that thrives on innovation.

Heinz and Christof Gebhart, Mobile Data Recorder Co-Creator at BMW Group, dreamed of adapting each development car to automatically transmit data over a cellular connection for faster aggregation and analysis. The primary goal was to get all the data out of the cars faster and provide it right away to BMW Group engineers so they could work together.

Faster data handling with an Azure cloud platform

Together, Gebhart and Heinz led a team that invented the mobile data recorder (MDR) system. Gebhart explains, “The MDR is a sophisticated IoT device that’s integrated into the car’s digital nervous system.” The device, about the size of a hardcover book, rests in the co-driver’s legroom and is connected to the development car’s network.The system has access to and can manage all electronic controllers and components in the vehicle. The MDR monitors and recordsmore than 10,000 signals twice per second and transmits data within seconds to a centralized cloud back end.

Heinz says it didn’t take long for the team to conclude they needed a cloud solution for their MDR system. “We started with local servers and faced various challenges, such as handling the pipelines and the huge amount of data,” he says. “The complexity and effort were too high, and scalability was only possible with a cloud solution.” They selected the Microsoft Azure cloud platform with Azure AI services,Azure App Service, and Azure Kubernetes Service (AKS) to bring their development data infrastructure to the next level.

“For our time-series analytics, Azure Data Explorer was the only way to go,” Heinz notes.

From the Azure platform, the team created an architecture and applications that would deliver faster, more efficient, and more comprehensive analysis. Production telemetry, stored in Azure Data Explorer, was initially accessed through Kusto Query Language (KQL) queries. The team developed back-end services and built the user interface with Vue.js. Both front-end and back-end services are managed as microservices on AKS, providing a responsive, interactive user experience. The deployed architecture also includes Azure IoT Hub and Azure Databricks.

Initially, mobile data recorders were installed in approximately 50 cars. The number of cars involved grew continuously, as did connections with other BMW Group data sets. The team also began two-way transmissions, sending configuration changes to the cars in real time. Early adopters helped expand uses for the data and identify how to perfect the platform and increase its usability.

Democratizing data with Azure AI

According to Heinz, the potential for AI quickly became apparent. “We want engineers without much KQL programming knowledge to be able to use our data. So, we connected with our teammates from Microsoft to see how we could adapt a generative AI chat experience to create KQL queries,” he says.

Using Azure AI Foundry and Azure OpenAI Service, BMW Group created an MDR copilot fueled by GPT-4o and accessed via web application. Engineers can now chat with the interface using natural language, and the MDR copilot converts the conversations into KQL queries, easing access to technical insights.

The copilot can also draw on a knowledge base created from expert interviews to answer more general questions about the development cars, data recorders, or related processes. Using Azure AI prompt flow, the team was able to rapidly create and test copilot prototypes and share them with management. Prompt flow also enabled one-click deployment to managed online endpoints, helping ensure seamless integration between the development and production environments and enabling A/B testing to improve responsiveness.

Copilot user interaction logging, conversations, and feedback and the general MDR knowledge base are reliably stored in Azure Database for PostgreSQL to facilitate the machine learning that builds the copilot’s efficiency over time. Azure Database for PostgreSQL also storesvectors, including KQL shots and the copilot’s Retrieval-Augmented Generation (RAG) chat patterns.

“The Azure AI works smoothly, significantly lowers the entry barrier, and will open up a much wider range of applications,” says Heinz.

The team also uses Microsoft Power BI to help democratize and visualize the data for decision-makers in all business roles. “We can put very complex raw data into an understandable and comprehensive web interface so many BMW Group employees who aren’t engineers are also able to access it,” Gebhart explains. “Azure is the turbocharger for delivering the right data to the right person on a large scale.”

Making auto innovation more agile

Moving from on-premises tools to a cloud-based system with faster data management helps engineers troubleshoot in real time. The vehicle data covered by the system has doubled, and data delivery and analysis happen 10 times faster. That reduces the lead time for insights from days to hours or sometimes minutes.

Saving time also decreases the cost per outcome. This enables engineers to send configuration changes back to development vehicles for additional testing, creating faster feedback loops and more agile development.

“The MDR system has been critical for the development of all our cars since 2020,” says Gebhart. “It helps guarantee a high level of functionality and quality in the cars, and we’re speeding time to market even as the number of features dramatically increases.”

As thousands of BMW Group engineers take advantage of the increased efficiency, faster prototyping, and higher development quality provided by the MDR system, the development department continues to advance. It is exploringMicrosoft Fabric as an end-to-end analytics platform that could further unify and simplify data management and visualization.

Because automotive innovation, especially in the digital domain, constantly accelerates, Heinz appreciates that Azure solutions do, too. He notes, “There are always new and innovative Azure features—and our development process benefits strongly from that.”

Gebhart agrees, adding, “BMW has always blended luxury and performance with cutting-edge technologies. With Microsoft’s cloud and AI leadership, we could capture the potential of our MDR system to accelerate development of our innovative cars. The MDR, powered by Microsoft Azure, ensures BMW reliability and quality long before the cars hit the road.”