Ben Savage On Stage
summary
Claim: combining DevOps practices with ML delivers deployable, versioned models at edge scale and faster business recovery. Implementation uses K3s + Fleet + Flux with Kubeflow for CI/CD of models, deploying versioned ML pods to devices. A real deployment handles thousands of devices each with 3–32 cameras for supervised image detection, plus a retrieval-augmented generation pipeline for evolving docs. Results: models update remotely, inference scales via Kubernetes, costs adjust programmatically, and chaos testing plus infra-as-code (Terraform) improves resilience.
AI-written summary · about
chapters (12)
- 0:00Introduction and speaker intro
- 1:24Speaker background and demos
- 2:25Old school machine learning
- 4:12Generative AI explained
- 7:16Private vs public models
- 8:20AI-generated band experiment
- 10:44Real-world use cases
- 11:25IoT image detection case
- 13:33Documentation AI assistant
- 16:00DevOps vs SRE for leaders
- 17:25Business resilience with DevOps
- 20:32Q&A and closing
related talks
57mDevOps for Machine Learning
17mDevOps for Private AI: A Scalable Platform Using Kubeflow
46mDeploying scalable Machine Learning models with Kubernetes
29mTeam Teardown: MLOps in the Cloud for IoT
42mDeploying ML at Scale with Kubernetes and TFX
39mOperating Deep Learning Pipelines Anywhere Using Kubeflow
27mDataEngBytes 2023 - SYD-T1-10 - Kanishka Mohaia
39mDeploy your ML Workloads on Kubernetes with Kubeflow