Julia Furst Morgado - Disaster-Proofing AI Workloads on Kubernetes

Even though vector databases are critical infrastructure for AI applications, most teams only consider backup strategies after disaster strikes. Whether it’s a corrupted index, accidental deletion, or a ransomware attack, losing vector embeddings can cripple RAG systems and leave LLMs hallucinating or returning empty results. In this session, we’ll explore what actually works (and what doesn’t) when protecting high-dimensional data, walk through Kubernetes-native strategies for backing up and restoring vector databases, and show a real-world recovery scenario that proves resilience doesn't have to come at the cost of performance.