Machine learning requires a good data set. It requires someone to build a model to make use of that data and make reasonable predictions about something, such as how many visitors are likely to buy, how many applicants deserve credit, or who is most likely to commit fraud. This kind of capability has been mostly out of reach for smaller companies, but new tools from a variety of vendors are starting to change that.
Those vendors include all the big cloud infrastructure players like Amazon, Google, and Microsoft, along with SaaS vendors like Salesforce and Adobe, and a slew of smaller companies. They all want to make it easier for you to get involved with machine learning, even if you work at a smaller organization.
Companies of all sizes can benefit from automating tedious tasks through the use of machine learning. Sometimes it doesn’t even require a data scientist because the vendor is providing the capability for you. Whatever path you take, you should be thinking about how you can make machine learning work for your clients.
Machine learning is coming to your software
This month at AWS re:Invent, the enormous annual customer conference in Las Vegas, AWS announced a slew of new machine learning tools built around SageMaker, the company’s managed tool for building, training, and deploying machine learning models.
At re:Invent AWS announced new SageMaker Studio, which includes notebooks, monitoring, and AutoML for automatically generating models. Amazon says its flavor of AutoML, called AutoPilot, gives you more insight into the models than similar tools.
Amazon isn’t the only solution out there. Just this week, Boston-based DataRobot, a full-service tool for building machine learning models in an automated way, acquired Paxata, a data prep startup. Together, the two companies give you a fairly comprehensive solution from preparing the data, building, and testing the model and putting it into production.
Another startup that launched this week, Arthur, wants to help you manage those models once they are in production. While models often perform well under the ideal conditions of a lab, a lot can go wrong when a model goes out into the world — from bias to data drift and lots of other issues. Arthur wants to help you monitor your machine learning model in the same way a company like New Relic monitors your systems and reports, should issues happen.
These examples just scratch the surface of the range of tools out there. Some of them require you to use your own data and build your own models, but some just put the technology under the hood and take advantage of the capabilities without you needing to think about it.
Regardless of how you implement machine learning, you need to be thinking about it in 2020 because it’s going to give you and your clients a way to reduce manual tasks, do more with less, and help you work smarter.
Photo: Syda Productions / Shutterstock