Back
GPU Cluster Management: From Chaos to Control
Online Webinar | May 14, 2026 | 1:00 PM - 2:00 PM CT
Overview
GPU clusters are expensive, scarce, and notoriously difficult to manage. Between NVIDIA driver compatibility, CUDA version requirements, node affinity rules, and scheduling conflicts, GPU cluster management feels like a full-time job—because for many teams, it is.
This webinar cuts through the chaos. We'll share practical patterns for managing GPU clusters at scale with Codiac: provisioning GPU node pools, managing driver and CUDA versions, scheduling training jobs efficiently, and keeping costs under control when GPUs cost $3-30/hour per node.
What You'll Learn
- The unique challenges of GPU cluster management on Kubernetes
- How to provision and manage GPU node pools across clouds with Codiac
- Scheduling strategies for GPU workloads: training, inference, and batch jobs
- Cost control: When to scale up, when to scale down, and when to sleep
- Live demo: GPU cluster lifecycle management with Codiac
Location
Online (Zoom)
What to Expect
Practical GPU management patterns with live demos. Ideal for teams spending too much time (and money) on GPU infrastructure.
Speakers
- Ben Ghazi, Co-Founder, Codiac
Who Should Attend
- Infrastructure Engineers managing GPU clusters
- ML Engineers frustrated with GPU scheduling
- FinOps practitioners tracking GPU cloud spend
- Anyone running AI workloads on Kubernetes