3. Models in R, Python, and Excel are largely
incompatible with web/mobile apps and CRM systems.
Developers must convert models into the language
they use in production (e.g. Java, .NET, Ruby).
Porting models from one language to another is
impractical, error-prone, and time-consuming.
No easy way exists for data scientists to integrate
analytical work into other systems without developers.
Greg Lamp | Austin Ogilvie | May 2013
4. Solution
A web platform for running predictive models
in production applications.
Greg Lamp | Austin Ogilvie | May 2013
5. ŷhat gives data scientists
Packages for deploying predictive web services
using the tools they know: Excel, R, and Python
An integrated framework for comparing and
horse-racing candidate models
Simple API for accessing models from any other
software system via REST
Web-based GUI for administration, reporting and
measuring the efficacy of models
Greg Lamp | Austin Ogilvie | May 2013
6. Greg Lamp | Austin Ogilvie | May 2013
Deploy models written in R
Publish models in a manner immediately useful to
developers.
Deploy Excel WorkbooksDeploy models written in Scientific Python
Consume models from any programming
language to make real-time predictions
7. Sectors and Applications
● Credit & Finance
● Marketing (response rate, loyalty,
& behavioral modeling)
● Sales (lifetime value, forecasting)
● Recommender Systems
● Fraud & Risk Management
● Systems Anomaly Detection
● Healthcare
Current Reach
● Started 3 months ago as side project
● Full time after June 1st
● 30,000+ unique visitors / month
● 200+ beta requests / month
● 117 predictive models deployed
● ŷhat highlighted at PyCon Silicon Valley
● Strong open source & social interest
● pandasql: open source project with
5,900+ users
● R & Pandas, 10 R Packages (HN)
● Conversations on Twitter
Greg Lamp | Austin Ogilvie | May 2013