Aflevering 123: How Kubernetes and AI Are Helping Prevent Wildfires

In this episode of The Dutch Kubernetes Podcast, Ronald and Jan talk with Andrea Giardini, cloud native consultant and trainer, live from Dutch Cloud Native Day. Andrea shares his journey into cloud and Kubernetes and dives deep into a real-world use case where Kubernetes, data engineering, and AI are used to help prevent wildfires.

Andrea explains how his client Overstory uses satellite and aerial imagery to monitor vegetation near power lines. By combining geospatial data, machine learning models, and infrastructure data from energy providers, they can calculate risk profiles and alert operators before vegetation causes sparks or fires. Instead of reacting to disasters, the platform focuses on prevention.

From a technical perspective, Kubernetes plays a critical role. The workloads vary massively, ranging from small CPU-based tasks to extremely heavy jobs requiring dozens of CPUs, large amounts of memory, or GPUs. Kubernetes provides the flexibility to dynamically scale these workloads, spin resources up and down when needed, and keep costs under control.

The conversation also covers the data engineering workflow. JupyterHub is used extensively for data exploration, but Andrea explains why notebooks alone are not reliable for long-term, repeatable processing. Once experiments are validated, workflows are moved into reproducible Python pipelines using a cloud-native workflow orchestrator (Dagster), fully integrated with Kubernetes.

They further discuss handling large datasets in object storage, running different pipeline steps with different resource profiles, GPU scheduling, and improving developer experience with pull-request-based preview environments. The episode highlights how cloud native technologies are not just about infrastructure efficiency, but can have real-world impact on safety, sustainability, and climate-related challenges.

‍