Empowering AI-Driven Drug Discovery: Overcoming Challenges in Building a ML Platform on Kubernetes
Talkat KubeCon + CloudNativeCon Europe 2025 · on Friday, 2025-04-04· at 11:00· Level 1 | Hall Entrance S10 | Room A·
summary
Built a Kubernetes platform that sits next to the DGX superpod Geon (almost 200 H100 nodes, ranked 21st) and exposes GPU capacity via MIG slicing, NVLink/NVSwitch and InfiniBand Quantum2 interconnect for low-latency training, plus an inference cluster named escard. Uses RunAI as a self-service UI, Harbor as a local Harbor cache and JFrog Artifactory to avoid 20+ minute image pulls, and integrates workload managers (Slurm/queueing) for large HPC jobs while supporting AKS, Argo CD and hybrid bare‑metal clusters.
AI-written summary · about
chapters (8)
- 0:00AI in pharmaceutical research
- 4:07AI in drug discovery pipeline
- 8:22Researcher's AI user journey
- 11:10Geion AI supercomputer
- 16:09LLMs on Kubernetes
- 20:23Run:AI platform demo
- 22:55Future hybrid clusters
- 24:21Q&A
related talks
27mGen-AI at Scale: Simplifying Orchestration of Healthcare Applications Across Multi-Cluster Environme
30mDesigning a Hybrid AI Platform on Kubernetes
31mKubernetes For GPU Powered Machine Learning Workloads In Academia
27mHarnessing Kubernetes for Scalable AI/ML Workloads
21mAIaaS with Dell AI ready infrastructure
10mChallenges and Opportunities in Making AI Easy and Efficient with...
30mRun Your Own AI Cluster on a DGX Spark: Kubernetes, GPUs, and DRA
35mCase Study: AI-as-a-Service on Kubernetes at Scale and In Production