Salesforce published some of the technology behind its AI-powered Einstein on GitHib Thursday. It's an AutoML library "for building modular, reusable, strongly typed machine learning workflows on Spark with minimal hand tuning" called TransmogrifAI.
The library is written in Scala and runs on top of Apache Spark, according to the website. It helps to streamline developer productivity with automation through machine learning and to focus on an "API that enforces compile-time type-safety, modularity and reuse."
- Converting some of this machine learning code to open-source will help alleviate the stress of learning how to scale ML within constraints, according to Shubha Nabar, senior director of data science at Salesforce, reports GeekWire. The library will allow companies to hire developers to implement TransmogrifAI.
Data scientists and developers are some of the most in-demand professionals in tech today. Both jobs focus on improving different aspects of business operations, but AI and ML could greatly impact that.
Salesforce started to build Einstein AI with three initial challenges, including the need for customer-specific models, a variety of structured customer data and scalability, according to Leah McGuire, principal data scientist for Salesforce Einstein, in an emailed statement to CIO Dive.
ML processes were automated from end-to-end with TransmogrifAI "to deploy thousands of customized models" without compromising data privacy, she said. Companies that struggle to have data scientists within reach could use the TransmogrifAI for ML-related use cases.
The open-source library could allow companies to customize how they cater and respond to customers, Byron Matthews, president and CEO of Miller Heiman Group, told CIO Dive. As a result, sellers can evolve how they consult with prospects.
"Modern sellers aren't exclusively offering buyers information about their company's products and services," he said, they have to deliver "actionable insight" so they "pay attention when a solution is put forth."