Dev Ops meets Data Science Taking models from prototype to production with Docker
summary
Shows a working path from prototype model to production API using Docker, Kubernetes and Anaconda and reports concrete outcomes: Docker image deploy in 10 minutes, 20,000 calls/day at early scale, and measurable SLOs for uptime and latency. Describes the API architecture, monitoring and log replay for debugging, use of Kubernetes for replica, canary and auto‑scaling deployments and notes reproducible environments via Anaconda YAML reproducible env and canary deployment that let new models roll over with traffic shift.
AI-written summary · about
chapters (10)
- 0:00DevOps and data science
- 1:27Prototype to production
- 5:20Data product constraints
- 10:22Service level agreements and monitoring requirements
- 18:06Hosting and infrastructure
- 22:22Scalable data services
- 27:58DevOps tools for data science
- 31:55Docker for reproducibility
- 33:43Kubernetes for data services
- 39:23Q&A
related talks
59mDevOps for Data Science
32mDeploying Python models to production
39mFrom Data Science to Production - deploy, scale, enjoy!
30mDevOps For Data Teams
01h 02mDocker for Data Science: Reproducibility and Deployment
35mGetting Machine Learning Models Ready For Production Using Python
56mDevOps for Machine Learning
22mDevOps in Data Science: What Works and What Doesn’t