kentoh - Fotolia

Using ModelOps, a financial services company scales out

Using model operations (ModelOps), a fintech startup was able to scale up its model deployment quickly, while also maintaining model governance at scale.

Exos, a provider of institutional finance services and vendor of a platform for B2B institutional finance, doesn't have a big staff, but it's getting bigger. The privately held firm, founded in 2018, has about 65 employees, and about half of them data scientists and engineers.

Despites its several dozen technical employees based at company headquarters in New York, Exos found last year that it had trouble scaling up the AI and analytics models that help power its financial services platform as the company grows.

To meet its growth, Exos needed a ModelOps platform to deploy its models faster.


ModelOps, or model operations, is an approach to more quickly build and deploy analytics and AI models. It relies on AI to automate the deployment, monitoring and governance of models at scale.

It's mostly the same concept as DevOps, but rather than focus on application development, it focuses on AI and analytics.

"CIOs are demanding scale and governance in their AI initiatives," said Stu Bailey, CTO and founder of ModelOp, said. ModelOps, he added, provides that scale and governance.

ModelOp, founded in 2016, sells a ModelOps platform for enterprises of all sizes. The platform enables users to scale up models quickly, and then manage the full lifecycle of their models.

While ModelOps is still a fairly new concept, the startup faces competition from several other vendors, including major tech players such as SAS and IBM.

ModelOp sells a ModelOps platform to help enterprises scale and govern their AI models.

Choosing a platform

Exos took those competitors into consideration when searching for a ModelOps platform, but found that the platforms from the big vendors were too "heavyweight" for Exos' composable infrastructure, said Joe Squeri, founder and CTO/COO at Exos.

Exos uses a lot of Open Source software, Squeri said. "With what we needed, we didn't want to marry ourselves to a large enterprise-wide platform," he said.

"We needed a tool that could fit into a narrower space than the big heavyweight products that are out there" and could still meet Exos' model control and scaling needs, he continued.

ModelOp fit those needs, Squeri said.

As we scaled, we saw the product itself scale.
Joe SqueriFounder and CTO/COO, Exos

Once signing up with ModelOp, it took Exos three or four weeks to get its first model running on the platform, and several more weeks to get the rest of its initial models running.

"We've been scaling ever since," Squeri said.

ModelOp has further shortened the already small divide between Exos' data scientists and engineers, Squeri said, explaining that the platform has enabled significantly more fluid organization and flow between model development and model testing and deployment.

The platform boosted Exos' pace of model deployment and has lent itself well to ensuring their models follow the financial industry's sometimes strict regulations, he said.

Exos keeps in close contact with ModelOp, giving feedback about the still fairly new platform nearly daily, Squeri said.

At first, "we were kind of going in a pace where we were maturing our models and building more and more models, we would give feedback to the team and the team was able to incorporate feedback in a relatively short order," he said. "As we scaled, we saw the product itself scale."

Squeri said he still looks at competing products occasionally but has found that, for now, ModelOp offers the best ModelOps platform for Exos.

Dig Deeper on Enterprise applications of AI

Business Analytics
Data Management