Joseph Sirosh, Microsoft | BigData New York City 2014
summary
Argues for an App Store for models and shows a working route: publish cloud-hosted models as APIs into Azure Machine Learning and a commercial Marketplace, demonstrated by a recommendations API you integrate with a few lines of JavaScript that scales to live clicks and bills on a pay-per-click basis. Describes a factory model for many domain-specific APIs (retail forecasting, wine vs. book recommenders), launched in summer and meant to avoid bespoke hiring by commoditizing services.
AI-written summary · about
chapters (14)
- 0:00Introduction and welcome
- 1:04Keynote on data science economy
- 3:27Recommendations API marketplace
- 5:31Domain expertise in machine learning
- 7:16Data signals and forecasting models
- 8:42Microsoft's cloud transition
- 10:26Data scientist shortage and solutions
- 12:31Mass manufacturing of APIs
- 14:58Factory for intelligent services
- 16:21Human factors in big data
- 18:06Probability and decision-making
- 19:32Becoming a data scientist
- 21:39Bias and data science ethics
- 22:43Conclusion and Q&A
