by Noelle Sio Saldana
Principal Data Scientist at Pivotal
The success of a Data Science project is not simply the model fit or the accuracy of its predictions; it is whether those models are being leveraged to make smarter business decisions. Over the past few years, Pivotal’s Data Scientists have experimented with software development methods practiced and taught by their Pivotal Labs counterparts in engineering, design and product management. By reframing Data Science as building software and products instead of research, we found that we reaped similar benefits: shorter and more productive iterations, and clients who actually used the models that we built and skills we taught long after we left.
In this talk, we discuss how we have successfully (and maybe not as successfully) borrowed principles from practices like Lean and Agile to Data Science. Topics include:
Minimum Viable Product Models
Build-Measure-Learn instead of a silver bullet
Pair programming
Scrums and retrospectives
Practicing empathy instead of elitism