Testing AI Containers for Digital Twins in Science: A Cloud-HPC Workflow
Talkat KubeCon + CloudNativeCon Europe 2025 · on Friday, 2025-04-04· at 11:45· Level 1 | Hall Entrance S10 | Room B·
summary
CERN and INFN build a CI/CD pipeline that uses Dagger + k3s + interTwin’s interlink adapters to submit containerized tests to SLURM HPC (Vega) from GitHub Actions, converting images via Docker-to-Singularity conversion and returning results to GitHub and a CERN Harbor registry. The itwinAI Python library runs distributed PyTorch/Ray workloads and Ray Tune HPO; hyperparameter tuning yields a 75% validation loss reduction in a hydrology use case. Automated multi-node GPU tests actually run on HPC.
AI-written summary · about
chapters (7)
- 0:00Digital twins in hybrid cloud-HPC
- 3:59Unified digital twin platform challenges
- 7:13Interlink adapter for Kubernetes on HPC
- 9:00Dagger runtime for AI workflows
- 11:31Demo: Interlink and Dagger
- 25:40Capabilities and future steps
- 27:47Q&A
related talks
35mState of Containers and the Convergence of Hyperscale, BigData and HPC
36mAn almost perfect match: cloud native and scientific computing
46mWhen AI Meets High Performance Computing
27mHarnessing Kubernetes for Scalable AI/ML Workloads
28mEnd to end AI developer journey with containerized assets using Intel DevCatalog, Intel DevCloud
17mDevOps for Private AI: A Scalable Platform Using Kubeflow
41mPanel: AI Infra Best Practices: Enterprise Do’s and Don’ts
25mLeveraging AI Ecosystem for Portable and Sustainable GPU Kernels in HPC