DevOps for Data Science
summary
Argues that applying DevOps practices to data science—source control, continuous testing, docker containers—gives reproducible models, repeatable pipelines and A/B model rollouts. Recommends versioning experiments (notebooks are fine for exploration but extract testable Python), codifying training environments as infrastructure-as-code, automating repeated training, and monitoring model scoring in production. Cites State of DevOps results (elite teams deploy ~46× faster and restore service ~2,500× faster) and notes tooling gaps today; prototypes succeed when pipelines save artifacts and track canonical data or change logs.
AI-written summary · about
chapters (27)
- 0:00Introduction
- 1:57Who is Damian Brady
- 3:30Data Science and DevOps
- 5:37What is DevOps
- 6:42DevOps Metrics
- 8:51DevOps Process
- 10:29Data Science Definition
- 12:44Data Science Cycle
- 14:35Data Science Processes
- 16:52Data Science Life Cycles
- 18:48Languages and Tools
- 23:12Operationalization
- 23:52Source Control
- 25:58Refactor
- 27:26Infrastructure as Code
- 28:52Automation
- 30:55Continuous Integration
- 31:57Continuous Testing
- 32:57Continuous Delivery
- 37:50Summary
- 40:35Pipelines
- 42:18Shift Left
- 43:09Ignite
- 43:51Jupiter Notebooks
- 45:54Data Breach Notebooks
- 46:49Data Breach Starter Mix
- 55:34Data Types
related talks
30mDevOps For Data Teams
29mDevOps for Data Science? - automate the boring stuff and leverage the OSS ecosystem
56mDevOps for Machine Learning
29mBringing Decades of DevOps Experience To Deep Learning
59mDevOps for Machine Learning Projects
30mDevOps to DataOps
51mFrom DevOps for data to DataOps
18mFrom DevOps to DataOps - Cloud based Software Development