Data science platforms

• Kaggle
• DrivenData
  • Data science challenges for social good
• Dream Challenges
  • Challenges in systems biology and translation medicine
  • Founded and managed by Gustavo Stolovitzky
• OpenML
  • Open platform focused on ML, especially supervised learning
• LabBook
  • Project at IBM Research Almaden and U. Toronto
  • Kandogan et al, 2015: LabBook: Metadata-driven social collaborative data analysis (doi, pdf)

All the major cloud providers now offer hosted data science services, currently branded as “AI” (AWS ML , Google Cloud AI , MS Azure AI , IBM Watson Studio ). Other offerings in this crowded space include:

• Domino Data Lab
  • “System of record” for data science work
  • Collaboration and reproducibility (automatic tracking)
  • Mostly infrastructure for computing in the cloud (uses Docker)
• Cloudera Data Science Workbench
  • Similar in spirit to Domino
  • Originally acquired from Sense
• Datazar
  • Hosted datasets, and notebooks and console for R and Python
  • Some custom visualization tools
• Plotly
  • Mostly hosted plots and dashboards
  • Claims to be “a GitHub for data scientists”

• Agave Platform
  • “Science-as-a-service”
  • Web API, not application
• H20
  • Implementations of common stats/ML algorithms for Hadoop
  • Also H20 Flow , a web-based notebook-style environment (video)
  • Open source , including Flow

• Knowen
  • Part wiki (organized as a DAG)
  • Part collaborative editor with LaTeX support (cf. Overleaf )
  • No computational features
• Hypothes.is (HN )
  • Open annotation of arbitrary webpages
  • UX via browser extensions