From the Life of a Data Scientist
summary
Claim: productive data science means shipping models into operation not slides. Core specifics: teams at Blue Yonder build predictive applications as a service that return actions, combine machine learning, software engineering and domain expertise, and aim for automated decisions. Tools and tradeoffs: use Python / scikit-learn to avoid an R-to-C++ rewrite, require out‑of‑core and parallel algorithms for large data. Results and metrics: example lift shows 67% of churners in the top 10% targeted, and probabilistic forecasts + MAE-focused costs drive deployment choices.
AI-written summary · about
chapters (26)
- 0:00Introduction
- 1:13Hype cycle and data science
- 3:42Industries in predictive analytics
- 5:49From descriptive to prescriptive
- 6:34Data delivery models
- 8:20Core skills of a data scientist
- 11:10Randomness vs causal effects
- 12:36Bayesian statistics
- 15:02Correlation vs causality
- 16:46Dynamic pricing
- 17:51Data exploration
- 21:02Machine learning fundamentals
- 23:54Supervised vs unsupervised
- 24:36Neural networks
- 25:39Curse of dimensionality
- 27:53Model fitting workflow
- 30:02Probability distributions
- 32:06Cost functions
- 34:16Business cases
- 37:07Operationalizing models
- 37:54Python for data science
- 41:11Handling large datasets
- 43:37Data storytelling
- 45:27Telecom and retail examples
- 47:37Demand forecasting
- 49:44Conclusion and operationalization in data science
related talks
43mDomain-driven Data Science
29mData Science Best Practices : From Proof of Concepts to Production
42mHow to be a 10x Data Scientist
37mBridging the gap: from Data Science to service
41mForget becoming a Data Scientist, become a Machine Learning Engineer instead
40mBuilding Successful Data Science Projects
46mPanel Discussion: Business Side of Data Science
22mWhy is a good data scientist a package of professions?