By using artificial intelligence (AI) and machine learning (ML) to draw valuable insights from data that is collected in accordance with data privacy laws and customers' consent, the BMW Group can make the experience of owning and driving its vehicles even more intuitive.
The BMW Group assigned Thoughtworks and others to build a scalable, cost-efficient and future ready platform for AI-based connected services and products – the start of the BMW Connected AI Platform. This would be built on top of a microservices platform based on Kubernetes clusters in the AWS cloud where we also worked for the BMW Group.
To deploy AI applications across the BMW Group's global operations, the AI platform needed to account for multi-region compliance regulations. The platform also needed to offer a standardized, scalable way to support all current and future connected AI use cases. And to ensure cost-efficient AI deployments, enabling multi-tenancy, and portability between cloud providers were also key considerations.
Overcoming technical complexity
A multidisciplinary team explored the available tools and deployment options that could help navigate the technical complexity.
BMW Group's data scientists should be enabled to deploy the necessary infrastructure for AI use cases on demand. Several tools to handle Identity and Access Management, and machine learning workflow orchestration were assessed. After assessing all the options together with the BMW Group, Kubeflow was selected, an open-source cluster orchestration framework. Coupling Kubeflow with other use case-specific tools accounts for the variety of the BMW Group's numerous AI use cases.
Opening up a world of machine learning and AI use cases
After four months, the POC of the platform was ready to test. The BMW Connected AI Platform is now live with its pilot use case: using machine learning to provide proactive care needed by the car. Based on readings from the connected car, the system identifies upcoming vehicle maintenance needs. The BMW Group can notify customers to schedule a service appointment before a potential defect occurs, providing a better customer experience (in accordance with data privacy laws and customer consent).
The platform has expanded to support further "Smart Maintenance" use cases of various vehicle components. The journey has started to facilitate many more AI use cases to get into production, encompassing even more aspects of the company’s connected vehicles, including the BMW Intelligent Personal Assistant.
Enabling data scientists to increase speed to market
Data scientists don't have to worry about infrastructure aspects like persistent storage, identity, access, and infrastructure security — everything is built into the platform.
As a result, the teams are enabled to apply Continuous Delivery for Machine Learning (CD4ML) best practices. They can work with large datasets without experiencing slow-downs or bottlenecks, allowing them to iterate on models faster. Version tracking enables them to track changes to code and data over time, which is important for reproducing results and troubleshooting.
All of this contributes to a more robust and secure development pipeline, allowing teams to fail fast and bring their use cases to market more quickly. International compliance is also built in, making global deployment of AI use cases much simpler.