Union AI Raises $19.1M in Series A, Simplifies AI and Data Workflows with Flyte

Image credit: union eye

Union AI, an open source startup based in Bellevue, Washington that leverages a cloud-native automation platform to help enterprises build and orchestrate AI and data workflows, today announced a Series A from NEA and Nava. announced that it had raised $19.1 million in a round. Ventures. The company also announced the general availability of its fully managed Union Cloud service.

At Union’s core is Flyte, an open source tool for building production-grade workflow automation platforms focused on data, machine learning, and analytics stacks. The idea behind this platform was to build a single platform that the team could use to create his ETL pipelines and analytics workflows, as well as machine learning pipelines. There are other projects on the market that offer similar orchestration capabilities, but the idea here is to build a tool specifically built for the needs of machine learning teams.

Flyte was originally developed within Lyft. Ketan Umare, his CEO and co-founder of Union AI, developed his early machine learning-based ETA and traffic models for the company in 2016. At the time, Lyft had to piece together a variety of open source systems. We will mass-produce these models.

“We ran something, but behind the scenes there was a man behind the curtain working. It was happening, but it was a lot of work,” Umarre said. “What we learned was that other teams within the company were also struggling, and these teams were large teams. It’s a big problem, but what’s the root cause?They couldn’t deliver and they couldn’t clearly explain why they couldn’t deliver.It’s the infrastructure. It turned out to be a problem with

Image credit: union eye

So he started working with a small team to build infrastructure tools to help those teams build models and deploy them in production. But there has always been friction between his software engineers and machine learning experts. “The reason is that, at least in the way I have extracted it, I think software and machine learning systems or his AI products are inherently different creatures,” Umarre argued. In his view, software typically matures over time, while AI models tend to degrade. These models are often modified based on external factors over which the user has little control, he noted. “So we can’t use the same infrastructure as before. [software deployments],” He said.

At that point, the team decided to open source their work in the form of Flyte and work with other teams to build a more machine learning native platform.

As is often the case, Umare and four other members of the original Flyte team then decided to build a startup around these core ideas and the Flyte open source project, and formed Union AI. was launched in late 2020.

Flyte is currently used by companies like blackshark.ai, HBO, Intel, LinkedIn, Spotify, Stripe, Wolt, ZipRecruiter and more.

“The interesting thing about working with these big companies, what we’re doing with open source, is that we’re working on some of the biggest models on the platform. We know this works and we’ve been doing this for years, so we didn’t have to build anything in particular, we just needed to extend a few things,” Umarre said.

“Based on a single team, Flyte will dispatch 10x more offline training jobs, resulting in 5x more frequent model releases, with significant business benefits,” Stripe said. said Mick Jermsurawong, Machine Learning Infrastructure Engineer at . “I think the realization here is that ML productivity isn’t a nice to have, it’s really a business requirement.”

But the Union AI Platform isn’t just building a Flyte-as-a-service. The team also built Pandera, a framework for data testing, and Union ML, a framework that sits on top of Flyte and helps teams build and deploy models using their existing toolsets. bottom. Union Cloud combines all these elements and layers a suite of enterprise tools such as single sign-on on top.

“Machine learning, especially large-scale language models, poses major problems when it comes to privacy and information security. It is,” said NEA Venture Partner Greg Papadopoulos. “Combining the power of big models with rich enterprise data can be tricky. This is why we are so excited about the progress of the Union.AI team, first Flyte and now Union Cloud. One of the reasons: It’s exactly what people are asking for and a real differentiator: Let us harness the power of language models at scale while maintaining control and ownership of your data.”