Machine Learning Meets K8s
summary
Argues Kubernetes solves ML pain points with concrete mechanisms: dynamic GPU allocation and autoscaling, namespace resource quotas, persistent volumes for checkpoints, and YAML‑driven portability across cloud or on‑prem. Demonstrates a simple TensorFlow job YAML that uses training operators, an end‑to‑end pipeline via Cube flow pipelines, and deployment to a KF serving REST endpoint for inference. Cites finance fraud detection and medical imaging as working, production use cases.
AI-written summary · about
chapters (16)
- 0:00Introduction to the Session
- 0:11Agenda Overview
- 1:11The AI Revolution and Infrastructure Needs
- 2:09Understanding Kubernetes
- 4:25Machine Learning Lifecycle
- 5:46Challenges Without Kubernetes
- 7:42Why Use Kubernetes for Machine Learning?
- 11:06Core Concepts of Kubernetes
- 14:42Introduction to Kubeflow
- 15:49Kubeflow Components
- 20:09Best Practices for ML on Kubernetes
- 21:12Resource Allocation Strategies
- 22:19Security Considerations
- 23:20Real-World Use Cases
- 24:53Future Trends in ML on Kubernetes
- 25:29Conclusion and Contact Information
related talks
31mKubernetes for Machine Learning: Managing and Scaling AI Workloads
42mKubernetes for Machine Learning: Managing and Scaling AI Workloads
46mDeploying scalable Machine Learning models with Kubernetes
51mMachine Learning using Kubeflow and Kubernetes
22mThe Containerization of Machine Learning: TensorFlow: Kubernetes, and Kubeflow
46mMachine Learning with Kubernetes
15mKeynote: Machine Learning on Kubernetes Made Easy With Kubeflow
12mKeynote: Building Enterprise AI/ML Platform on Kubernetes