Data Science at PMI - The Tools of The Trade
summary
Thirty plus data scientists across Amsterdam Lausanne and Cracow build an in‑between platform (not pure cloud, not only Hadoop) that pushes models to production via docker data science labs, Bitbucket, j-rock artifactory scanned by Black Duck, Jenkins, Airflow and Cabana for monitoring. Templates and a cookie‑cutter enforce raw data + Dockerfile + code reproducibility, a three‑stage pipeline (phase zero → alpha → beta → industrialization) and a CI/CD flow so artifacts ship fast rather than lingering in notebooks.
AI-written summary · about
chapters (29)
- 0:00Introduction
- 1:54Data Products
- 2:57Outcome
- 3:38Culture Conflict
- 4:45Data Science Team
- 5:48Data Science Principles
- 7:41Todays Data Science
- 8:13Perception of Data Science
- 8:42Selftaught Data Scientists
- 9:15Google Rule of Machine Learning
- 10:12Data Science Infrastructure
- 11:13Data Sources
- 11:48Workflow
- 12:57Architecture
- 15:00Best Practices
- 15:16Docker Containers
- 16:33Project Templates
- 18:06Cookie Cutter
- 20:09Quote
- 20:38Continuous Integration
- 22:18Software Release Cycle
- 26:32Coding Standards
- 28:29Deployment
- 29:16Monitoring
- 30:23Airflow
- 31:23Best Practice Ambassador
- 32:05Technology Scouting
- 32:32Continuous Learning
- 33:02Conclusion
related talks
19mData science for real problems
19mDepend on Docker for Self-Managed Predictive Modeling
30mDevOps For Data Teams
29mData Science in Production
39mFrom Data Science to Production - deploy, scale, enjoy!
59mDevOps for Data Science
01h 02mDocker for Data Science: Reproducibility and Deployment
46mDev Ops meets Data Science Taking models from prototype to production with Docker