Sightwise

Background

Sightwise, a European AI company developing manufacturing applications relied on an on-prem Kubernetes cluster to power synthetic data generation and ML workloads. These processes required significant GPU resources for ray tracing, simulation, and model training. As workloads grew, maintaining on-prem infrastructure became increasingly complex and limited the team’s ability to scale GPU capacity quickly. The company needed a cloud platform that could support high-performance AI workloads while reducing operational overhead.

Challenges

Operational Maintenance Overhead: Self-hosted services such as S3-compatible object storage and GitLab CI/CD required constant management alongside on-prem hardware infrastructure.

High GPU Demand: Synthetic data generation and ray tracing workloads required powerful NVIDIA GPUs that were difficult to scale efficiently on-premise.

Limited Deployment Automation: The existing Kubernetes environment lacked automation and resilience, slowing deployments and ML experimentation.

Solutions Delivered

Zazmic designed and implemented a scalable AI infrastructure on Google Cloud:

GPU-Optimized GKE Clusters: Deployed a multi-node GKE cluster optimized for AI workloads using NVIDIA L4, RTX PRO 6000, and T4 GPUs to support compute-intensive tasks.

Lift-and-Shift Cloud Migration: Migrated Kubernetes workloads to Google Cloud with minimal application changes, enabling rapid adoption while preserving existing workflows.

Managed Data & Storage Services: Moved databases to Cloud SQL and object storage to Google Cloud Storage, eliminating the maintenance burden of self-hosted infrastructure.

GitOps CI/CD with ArgoCD: Implemented automated deployment pipelines using ArgoCD and Artifact Registry for consistent, repeatable releases.

Conclusion

Zazmic delivered a scalable Google Cloud platform that enables Sightwise to run GPU-intensive AI workloads efficiently. By combining GKE, managed cloud services, and automated deployments, the solution reduces operational complexity while providing the performance and flexibility needed for advanced manufacturing AI development.