Leading business collaboration platform deploys models faster with Verta MLOps

The Challenge

A leading business collaboration platform’s ML Engineering team regularly serves new machine learning models to production. Their lead ML Engineering Manager describes how processes prior to Verta were problematic.

No clear method for moving ML models to production

“In the beginning, serving a model to production was a big question each time,” he remembers. Having to “reinvent the wheel” each time the team served a model led to significant inefficiencies.

Internal solutions were limited by engineering capacity

ML Engineers built a simple solution to serve models to production, but this pursuit was constrained by time. “Because we're not a huge team and we didn’t have the time to invest to make it robust or easy to use,” recalls the ML Engineering team lead, “so we had to go out and look for a better solution.”

Other solutions didn’t work with existing infrastructure

“We have a lot of security requirements, and we have to be very careful with our customer's data,” he explains. “We can't ship data outside of our VPC, so an externally hosted solution would just never work. It had to run on our infrastructure.”

The Solution

Verta provided a unified MLOps platform that addresses major ML Engineering deployment and operations issues.

Faster implementation of new ML models

“The timeline for developing a new machine learning application involves a bunch of steps,” the customer explains. "If we cut down on that timeline, we increase our iterations. And that’s what Verta has let us do.”

Verta works with existing infrastructure

Verta allows the customer to maintain strict data privacy and security controls. “Verta is basically built on our infrastructure,” says the customer. “From our point of view, it's running within our VPC on our infrastructure in a way that satisfies our security requirements. No other vendor could do this.”

The Solution

More time for engineers

Most importantly, Verta has freed ML Engineers to focus on solving more strategic business problems. “The big thing for me as an engineering manager is I want my engineers working on things where their effort is really differentiated. They should be doing the things only an engineer can do,” he asserted. “So, taking models to production have evolved — from creating a bespoke solution each time to now, with Verta, it's just a solved problem.”