DataOps and MLOps: NLP Development and Deployment in Kubernetes made easy
summary
A Kubernetes-based DataOps/MLOps pipeline is built and running: code in GitHub Codespaces produces ephemeral images for dev testing, Argo manifests trigger GitOps via Flux and Terraform on AWS, training logs go to MLflow, and models deploy behind Seldon for API serving; vcluster and Carpenter isolate workloads and optimize node selection. Monitoring uses Evidently plus Datadog; TF‑IDF (NLP) and MAE (time series) metrics signal drift. The hybrid setup works end‑to‑end and removes repeated manual deploy steps.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:13What are DataOps and MLOps
- 3:24Why these disciplines matter
- 7:22Existing solutions and challenges
- 10:08Custom and hybrid approaches
- 11:14Key tools for ML lifecycle
- 14:31Common challenges in projects
- 16:17Use case overview
- 17:20Platform requirements
- 19:29Kubernetes and tooling setup
- 21:14Orchestration and ML tools
- 22:38Development workflow demo
- 24:45Deployment flow explained
- 25:28Key advices for success
- 26:51Future trends in operations
- 30:04Closing remarks and Q&A
related talks
29mBosch DataOps: NLP Development on Kubernetes
26mKubernetes and MLOps for Scalable and Reproducible Generative AI
53mExperience with an end-to-end MLOps pipeline: Practical insights and 3D architectural perspectives
29mDevOps for Data Science? - automate the boring stuff and leverage the OSS ecosystem
36mIs the MLOps tooling the same as the DevOps tooling?
44mDevOps for Machine Learning - why is it different?
45mArchitecture of an NLP Deployment
29mTeam Teardown: MLOps in the Cloud for IoT