Union.ai Raises $38.1M Series A to Expand AI Development Infrastructure

Union.ai has closed a $38.1 million Series A round, more than doubling the $19.1 million portion it previously announced. The round was led by existing investor New Enterprise Associates (NEA), with participation from Nava Ventures and new investor Mozilla Ventures. The capital will be used to accelerate product development, grow engineering and field teams, and deepen investment in the company’s open-source ecosystem.

The raise lands at a moment when AI orchestration is moving from niche concern to frontline engineering priority. As more organizations attempt to productionize AI systems, the gap between experimentation and reliable deployment has become harder to ignore.

From Open Source Roots to Enterprise Platform

Founded in 2020, Union.ai began as the enterprise platform built around Flyte, an open-source AI orchestration framework originally developed at Lyft. Over time, the company expanded beyond orchestration to offer a broader AI development platform spanning model training, inference, and observability.

Today, Union.ai positions itself as the development layer of the modern AI stack — focused on helping teams manage dynamic, non-deterministic workflows that traditional data tooling was never designed to support.

Unlike conventional software pipelines, AI workflows often involve runtime decision-making, long-running processes, and shifting dependencies between data, models, and compute. That complexity has created demand for infrastructure purpose-built for AI rather than adapted from legacy systems.

Why AI Infrastructure Needs a Rethink

Engineering teams are discovering that classic CI/CD pipelines and deterministic workflow tools struggle under the demands of agentic and model-driven systems. AI applications must handle branching logic at runtime, recover gracefully from failures, and version data and models automatically — often at scale.

Union.ai’s platform focuses on solving these operational challenges by enabling:

• Pure Python workflow authoring with local testing before cloud deployment

• Live remote debugging and improved failure visibility

• Fully dynamic workflows for agentic AI applications

• Scalable, long-running executions with automated versioning

• Crash-resilient pipelines with built-in retries and caching

The goal is less about flashy demos and more about making AI systems durable in production environments.

Continued Commitment to Open Source

Open source remains central to Union.ai’s strategy. The company recently introduced Flyte 2, the next generation of its orchestration framework, which has surpassed tens of millions of downloads. Its broader open-source portfolio also includes Pandera, a data validation framework widely adopted across the Python ecosystem.

Mozilla Ventures’ participation in the round reflects alignment around open AI infrastructure — particularly at a time when questions of transparency, portability, and developer control are becoming more prominent in the AI ecosystem.

Open Source and the Long-Term Implications for AI Infrastructure

Union.ai’s trajectory is closely tied to its open-source foundation. Flyte, along with projects like Pandera, has become embedded in thousands of engineering environments not because of aggressive commercialization, but because developers adopted the tooling organically. That adoption model matters. In AI infrastructure, open source often acts as both distribution channel and trust layer — enabling teams to inspect, modify, and extend the systems they rely on.

As AI systems become more central to business operations, the implications of open infrastructure grow more significant. Organizations are increasingly wary of locking mission-critical workflows into opaque platforms. Open-source AI orchestration offers portability, community validation, and the ability to evolve alongside fast-moving model ecosystems.

In practical terms, infrastructure like this could influence how the next generation of AI systems is built. Rather than monolithic platforms controlling the full stack, the future may favor composable architectures — where orchestration, model management, validation, and observability remain modular and interoperable.

If that model holds, companies that cultivate strong open ecosystems while offering enterprise-grade reliability may help define the technical foundation of production AI. In that sense, the broader impact of Union.ai’s work extends beyond its commercial platform — it touches the architecture choices that will shape how AI is built, governed, and scaled in the years ahead.