Follow Datanami:
March 17, 2017

Data Platform Speeds Model Deployment

(Mini bear/Shutterstock)

Data science platform startup Domino Data Lab, which launched its first product in 2014, rolled out its first upgrade this week designed to speed the deployment of predictive and other models along with the addition of an underlying model deployment architecture based on the Kubernetes cluster orchestrator.

The San Francisco-based company said its Domino 2.0 platform includes a reworking of its “API endpoints” functionality designed to allow data scientists to deploy models as REST APIs. That new architecture is intended to enable “web-scale” model deployment with lower latencies and the ability to run model variations, the company said.

The new model deployment capability “addresses a clear need of almost every data science organization we talk to,” Nick Elprin, Domino’s CEO and co-founder, noted in a blog post announcing the upgraded data science platform.

Elprin said the model deployment upgrade would help data science teams ship Python and R models to production faster as “high-availability API endpoints.” The upgraded platform is based on Kubernetes, which is emerging as the go-to cluster orchestrator for application containers. The Domino platform runs Docker containers as an abstraction layer and a buffer between the data scientists and the models, the company told Datanami last month.

Among other advantages, the reduced latency “allows data science teams to work with their engineering counterparts to embed data science results into business-critical processes rapidly without being forced to ‘dumb down’ models or retranslate them into another language,” Elprin explained.

The upgraded platform also supports A/B testing that would allow, for example, comparisons of competing recommendation engines built in Python and R languages. “Software development has long-embraced rapid experimentation in production and there’s no reason data science should be any different,” asserted Elprin.

The data science platform also indicates how application container technologies are making inroads as data science platforms emerge. Domino Data Lab noted that its upgraded platform leverages Docker as a way to boost collaboration between data scientists and business users.

Other new features include a new user interface designed to streamline workflows by allowing data scientists to review past results and track progress on research projects. The upgrade also includes automated reports as a way to keep business users in the loop as models are retrained to account for new data.

The company said Domino 2.0 is available now to beta customers, and will be generally available later this spring.

The company announced a $10.5 million funding round in November 2016 led by Sequoia Capital. The startup’s early customers include pharmaceutical and life sciences companies along with insurers and financial services firms. In announcing the funding round, Elprin said the startup wanted to use the cash infusion to “throw gas on the fire” as platforms emerge to deploy data science projects in one place.

Recent items:

A Platform Approach to Data Science Operationalization

Why You Need a Data Science Platform