Scaling Data Science
summary
Bolt builds a centralized feature store with SQL/JSON feature definitions, daily snapshots and data‑quality checks, plus Redis for low‑latency lookups, Elastic for reverse queries, Kafka/S3/Redshift for pipelines, Airflow for orchestration and SageMaker for early training. The result is concrete: new features go from pull request to production overnight, prototypes appear in two hours, and heavy jobs reduced job time by 80%. The approach emphasizes reusable features, measurable cycle‑time savings, and real incidents caught by data checks.
AI-written summary · about
chapters (24)
- 0:00Introduction
- 1:25Have you used Bolt
- 2:25Why is data science important
- 3:17Challenges
- 5:08Scaling Data Science
- 9:59Drawing Tutorials
- 10:21Machine Learning
- 11:44Scale Data Science
- 13:23Data Warehouse
- 14:20Preprocessing
- 15:04Model Lifecycle
- 16:22Most Important Thing in Machine Learning
- 17:00Feature Engineering
- 18:02Feature Store
- 19:01Machine Learning in Production
- 22:16Data Scientist Language
- 24:58Technical Changes
- 26:02New Challenge
- 26:44Cues
- 30:31New Reality
- 31:20Key takeaways
- 33:10Next week
- 33:32About Bolt
- 34:17Questions
related talks
28mData VS Engineers: a long lasting story
33mFeathr, an Open Source, Battletested, Scalable and Enterprise-Grade Feature Store
30mHow Feature Stores Enable Operational ML
58mPrepare data and model features for ML with ease, speed, and accuracy
25mBuilding A Feature Store For Hyper Growth
29mData foundation for analytics and AI
33mEvolution and Scaling of Feature Store at Uber
30mAmazon SageMaker Feature Store: Store, discover, & share features for ML apps