Embedding AutoML Into Larger Development Projects

One of the key advantages of AutoML is the ability to quickly build accurate predictive models for "microdecisions" that an enterprise has. Most organizations use machine learning today primarily for their most important decisions. With AutoML even small events can be predicted based on recently gathered data, without the typical multiweek data science project.

Building AutoML training into your applications, not just executing predictions but building and training models based on dynamically assembled data can make machine learning a pervasive way to do business.

All AutoML tools have web service interfaces that make this easy to do. At this meetup, we will show how to assemble a training dataset CSV dynamically from disparate sources, execute experiments, promote a winning pipeline and produce predictions from the promoted pipeline using new data. This procedure will work with all major open source AutoML products: Auger.AI, H20, TPOT, and AutoSKLearn.

We'll provide some small bites and pizza, plus water and soft drinks. A no-host bar is available for beer and wine.

Schedule:
6:30-7pm - Sign-in, eat/drink & get to know each other
7-7:50pm - Presentation