Mastering the DevOps Machine Learning Pipeline
summary
Describes a production pipeline that runs K3s on devices and in cloud, uses Fleet and Flux to deploy versioned pods with distinct ML models across thousands of IoT units each with 3–32 cameras, and hosts open‑source models to avoid public LLM drift and training leakage. Shows retrieval‑augmented generation for searchable onboarding docs, warns GPU build costs, recommends infrastructure as code with Terraform, automated tests (real deployments beat 25% coverage), and chaos engineering with Gremlin or Chaos Monkey for resilience.
AI-written summary · about
chapters (11)
- 0:00Introduction to talk
- 1:06Speaker background
- 2:11Old school machine learning
- 3:57Generative AI explained
- 5:23Data privacy in ML
- 7:48AI-generated band experiment
- 10:53Real-world IoT use case
- 13:00Documentation with generative AI
- 15:46DevOps vs SRE
- 17:10Business continuity with DevOps
- 19:58Conclusion and Q&A
related talks
57mDevOps for Machine Learning
56mDevOps for Machine Learning
29mTeam Teardown: MLOps in the Cloud for IoT
24mFrom Model to Production: Automating the ML Lifecycle with CI/CD
44mDevOps for Machine Learning - why is it different?
39mOperating Deep Learning Pipelines Anywhere Using Kubeflow
17mDevOps for Private AI: A Scalable Platform Using Kubeflow
05mCI/CD Pipelines and Machine Learning