Enabling collaboration among competing divisions

Competing subsidiaries mindful of sharing data

While scale certainly has its benefits, it also comes with its own set of challenges. As a high-tech engineering company with clients in manufacturing, mining and infrastructure, and 41,000 employees spread across 170 countries, Sandvik knows all about those benefits and challenges.

Being a manufacturer with thousands of products being delivered to different customers all over the world, Sandvik has leaned heavily into machine learning to know what product they are going to make, where they are going to make it and to what country the product will be exported. William Smedberg, AI Portfolio Manager at Sandvik, is tasked with the delivery of AI capabilities within the whole Sandvik Group. As a large global corporation with multiple divisions, a big challenge for Sandvik lies in consolidating their data and being able to share insights and projects across divisional borders. 

“Each of our business units operates autonomously and doesn’t typically share data with others, sometimes because they are direct competitors,” William said. This structure led to data silos and inefficiencies, as innovations and insights generated by one division could not easily be applied across the organization. The decentralized nature of data made collaboration difficult, preventing Sandvik from leveraging their full potential.

Unified data, accelerated innovation

Databricks provided a solution to these issues by enabling centralized data connectivity, without disrupting the autonomy of each business unit. “With Databricks Data Intelligence Platform, divisions can operate independently while remaining connected in the background. That has two great advantages: more efficiency within each individual data team but also the ability to take something one team has built and use it in other parts of the organization,” William said. This allows the seamless transfer of AI models, code and other data assets across the organization, fostering greater collaboration and efficiency.

“The packaging and distribution of data, code, models and other assets can then be used by other teams, from raw data to serving of the final product. Using serverless compute, both for SQL and Python, both the developer and consumer experience becomes seamless and without wait times. Developers can simply take prepackaged assets and deploy them on their own data in a matter of seconds,” William continued.

A key use case at Sandvik has also involved leveraging AI for real-time monitoring of different types of mining equipment; for example, drilling machines. The Databricks Platform allows the company to analyze sensor data from machines in operation, helping customers optimize maintenance schedules and minimize downtime. “If a sensor detects an issue, AI models can immediately provide a solution to the operator, based on a vast database of machine manuals and historical data,” William noted. This application of AI has the potential to drastically reduce operational costs for customers by preventing equipment failures.

Lastly, Sandvik has also leveraged Databricks Data Intelligence Platform to streamline the development process for their engineers. Through the platform’s Unity Catalog, developers gained visibility across multiple environments, making it easier to manage data workflows and test new AI models. This reduced the need for manual intervention and helped accelerate innovation.