State of Data Science & Machine Learning
summary
Claims that packaging and tooling drove Python's adoption for data science supported by concrete projects and usage numbers: a self-contained Conda packaging manager and Anaconda distribution that remove install pain, acceleration via Numba, scale via Dask, visualization via DataShader and cataloging via Intake, weekly uniques ~1M and trailing twelve months ~16M. Roadtested products exist, enterprises adopt an enterprise platform built on Kubernetes and Docker, and the built tools demonstrably work and shift where engineering effort must go.
AI-written summary · about
chapters (36)
- 0:00Introduction
- 0:32Peters background
- 2:33Anaconda
- 5:52The Journey
- 10:54Continuum Analytics
- 12:17Python Packaging
- 13:07aconda
- 13:38Conda
- 14:19Adoption Curve
- 15:58Packaging
- 17:13Numba
- 17:44Task
- 18:15Large Data Visualization
- 18:52Intake
- 19:22Lessons learned
- 20:00What comes next
- 20:38Transformation
- 21:05Challenges
- 22:07Human Relations
- 23:15Engineering Questions
- 23:34Isaac Newton
- 23:52The Long Shadow
- 25:09Pragmatic Programmer
- 25:56The Clean Coder
- 26:41Trust
- 27:38Humility
- 30:32Whats next
- 30:44Hardware innovation
- 31:23New chip architectures
- 31:51Deep learning
- 32:29Innovation space
- 32:52Outsourcing innovation
- 33:28Predict button
- 35:18Machine learning vs data science
- 36:23Authorized computing
- 36:40Service computing
