Typedef emerges from stealth with $5.5M to turn AI prototypes into production-grade pipelines

Typedef, an AI startup focused on turning AI prototypes into production workloads, just came out of stealth with $5.5 million in seed funding. The round was led by Pear VC, with backing from Verissimo Ventures, Monochrome Ventures, Tokyo Black, and a group of angel investors.

The company is run by co-founders Yoni Michael and Kostas Pardalis, both of whom have deep roots in data infrastructure. Michael previously sold his data center analytics startup, Coolan, to Salesforce in 2016. The two call themselves “data nerds” and are setting their sights on the $200 billion AI infrastructure market.

“We’re also extremely proud to announce our $5.5M seed round led by Pear VC, with support from Verissimo Ventures, Monochrome, Tokyo Black, and an exceptional group angels. We’re incredibly fortunate to have such a sharp and supportive group of backers who share our belief that AI infrastructure needs to be reimagined from the ground up,” Typedef said in a blog post.

Their pitch is simple: too many AI projects never make it past the prototype stage. Typedef was built from scratch to handle the messy reality of production AI—running large models, managing unstructured data, and dealing with workloads that don’t behave predictably.

“Inference is the new transform,” the company says, referring to the shift from training models to actually running them at scale. The team has experience leading infrastructure efforts at companies like Salesforce, Tecton, and Starburst Data, and they’ve poured that into Typedef—a tool meant to help teams move beyond duct-taped AI setups that break under real use.

“Typedef is ushering in the new era of AI infrastructure where model training has given way to inference and where teams can build reliable, scalable, and cost-effective Large Language Model (LLM) workloads without the complexity or strain of managing infrastructure,” said Arash Afrakhteh, partner at Pear VC.

Most AI projects never leave the prototype phase

It’s not just a hunch. AI pilot projects are stalling everywhere. According to a survey by Informatica, 93% of U.S. data leaders are increasing GenAI budgets next year—but 67% say they haven’t even been able to move half of their pilots into production. Some estimates say 87% of AI projects fail to scale.

“There’s no clear path to production,” said Michael. “Legacy systems weren’t built for LLMs or unstructured data. So people stitch together outdated tools, custom scripts, and hope for the best. But the result is usually something that’s brittle and unreliable.”

Typedef is trying to fix that. It lets teams build, test, and deploy AI pipelines—workflows that use LLMs to process and analyze data—with far less friction. It was built to be deterministic on top of non-deterministic models, which means teams can trust what they ship.

What Typedef Does

At its core, Typedef helps developers run scalable LLM pipelines for things like semantic analysis without worrying about all the messy pieces like token limits, context windows, or data chunking. Its interface is simple, with APIs and relational models that engineers are already used to.

It’s fully serverless—no infrastructure setup, no provisioning, no config files. You download the open-source client library, hook it into your data sources, and start building pipelines with a few lines of code.

“Teams want the same reliability from AI pipelines that they’ve had with traditional data pipelines,” said Pardalis. “They want to run analytical workloads using AI, extract insight from their own data, and do it without fighting the tools every step of the way.”

Early Traction From Real Customers

Matic, an insurance-tech company that works with over 70 carriers, is already using Typedef in production. They built AI workflows to process policy documents and customer support transcripts. The result: faster processing, fewer human errors, lower compliance risks.

“Typedef lets us build and deploy semantic extraction pipelines across thousands of policies and transcripts in days, not months,” said Lee Maliniak, Chief Product Officer at Matic.

What’s Next For Typedef?

Typedef is aiming to become the go-to infrastructure layer for companies trying to operationalize AI, especially when it involves large models, unstructured data, or complex workflows. The startup believes its stripped-down, production-ready approach is exactly what companies need to finally break out of AI pilot purgatory.