Data science & Agility at Springer Nature
summary
Springer Nature moves from ad hoc dumps to production-ready pipelines by making source teams publish data as a product with metadata, contracts and observability; lineage shows who uses what across 40+ products. The workflow shifts cleaning and transforms left so APIs serve precomputed outputs and recommendations (recomputed hourly) instead of running heavy models at request time. Implementation uses Google Cloud tooling, notably BigQuery and dbt, and continuous quality checks to reclaim trust in metrics and reduce runtime compute.
AI-written summary · about
chapters (8)
- 0:00Data science & agility intro
- 1:47Three challenges in delivery
- 8:53Shift left as vaccine
- 11:44Data as product benefits
- 21:35Incentives for data products
- 27:16Delivery models & MLOps
- 33:20Precompute ML for efficiency
- 34:02Summary & takeaways
related talks
22mBuild Smarter, Don't Just Build
44mPeople, Technology, Culture: How to make Data Science work
29mData Science in Production
24mHuman Friendly, Production Ready Data Science Stack
19mA world of data
46mData-Driven Transformation: Organizational Changes and Leadership Insights
38mScaling Data Science
07mFrom DevOps to DataOps