Building Secure MLOps Pipelines With KitOps + Argo Workflows
summary
Shows packaging a sklearn sentiment model (pickle) into a CNCF Sandbox model kit using the kit pack command, pushing the immutable OCI artifact to a registry, then running single and batch inference by unpacking that kit inside an Argo workflow DAG. Demo produces expected negative sentiment log output, batch run spawns parallel pods and aggregates results, supports GPU pods, GitOps-driven registries and attestation for traceability and reproducible, auditable deployments.
AI-written summary · about
chapters (11)
- 0:00Secure MLOps introduction
- 1:06Production ML challenges
- 6:08KitOps for ML packaging
- 12:11Argo Workflows overview
- 13:34Packaging ML with KitOps
- 15:40Single inference workflow
- 18:50Batch inference with DAGs
- 21:20Live demo: single inference
- 24:12Advanced Argo patterns
- 25:36Recap and benefits
- 27:01Q&A
related talks
32mBridging DevOps and MLOps: Unifying Pipelines with KitOps and GitOps
18mBaby Sitting Your AI Models in Production? STOP!
25mBuilding MLOps POCs and Sandbox Environments Using k3s and Argo
36mDataOps and MLOps: NLP Development and Deployment in Kubernetes made easy
28mScaling MLOps on Kubernetes with Amazon SageMaker Operators
23mBringing Container-Native Simplicity to AI/ML
24mDemo of Kubernetes at the Edge for AI Model Deployments
22mBuilding a MLOps Platform with Kubeflow on GKE