Typedef is a new AI startup solving an age-old problem: data optimization. Traditional data engines were built for rows and columns, not the messy, multi-modal inputs of modern AI. Typedef has emerged with a new AI data infrastructure built from the ground up to help developers and software engineers better manage modern workloads. In other words, the company’s technology can help companies transform AI prototypes into production-ready, scalable workloads that quickly produce business value.
Typedef Emerges Out of Stealth Mode
In June 2025, Typedef unveiled its AI data engine to the world after raising a whopping $5.5 million in seed funding from investors who believe in the company’s vision. The largest early-stage investor is the seed-stage venture capitalist firm Pear VC, with notable participation from Tokyo Black, Monochrome Ventures, Verissimo Ventures, and numerous angel investors.
Arash Afrakhteh, who is a partner at Pear VC, spoke highly of the Typedef team in a recent interview. He believes Typerdef is paving the way for a new era of AI data infrastructure where model-training has given way to inference. To expedite AI projects, developers want to use familiar tooling and query engines that can handle structured and unstructured data with equal precision and predictability.
The Founders
Yoni Michael and Kostas Pardalis are the two serial entrepreneurs who founded Typerdef. They refer to themselves as “data nerds,” but they are two skilled data infrastructure engineers with a long track record of success. Part of this success included managing data infrastructure teams at major cloud-first companies, such as Starburst Data, Salesforce, and Tecton.
Michael founded and ran a data center analytics company called Coolan, a small startup that managed fleets of data centers. In 2016, Michael sold Coolan to Salesforce for an undisclosed amount. Within years after the sale, the AI boom hit the IT world and inspired the two data nerds to grab their share of the AI infrastructure market, worth an estimated $200 billion.
The vast experience and skills acquired from leading these teams gave Michael and Pardalis the confidence to start Typerdef. Their mission is to bring a new type of AI data processing engine built from the ground up to run production AI workloads at scale – ultimately operationalizing AI across data stacks and the enterprise.
Why Many AI Projects Never Make it to the Production Phase
Did you know that nearly 87% of all AI projects fail to make it past the pilot phase? This is the phase where a trial of a prototype AI development is released to a select number of developers for testing purposes. The developers use the pilot phase to identify potential issues with the software and gather the necessary data to make the required improvements.
Most new AI projects in the tech industry involve Generative AI, a type of AI that generates new content by analyzing existing data. According to the CDO Insights Survey from Informatica, approximately 93% of American data leaders plan to make larger investments in Generative AI technologies in 2025. Unfortunately, the survey also revealed that a stunning 67% of organizations have been unsuccessful in transitioning at least 50% of their Generative AI projects from the pilot phase to the production phase.
These projects are stuck in what is known as “pilot paralysis.” Tech researchers use the term “pilot paralysis” to describe the epidemic of most enterprise AI projects failing to progress past the pilot phase. One research report even suggested that 87% of these AI projects never reach the production phase.
Typedef co-founder Yoni Michael suggested the cause of this alarming failure rate is that legacy data platforms were not designed to handle unstructured data, inference, and large language models (LLMs) forcing developers to focus on managing complex infrastructure with brittle data-processing pipelines that are unreliable and don’t scale.
Typedef Provides the AI Data Infrastructure for Modern AI Workloads
Typedef makes it easier than ever to run scalable LLMs at significantly lower costs and operational overhead. Data engineers and software developers aren’t accustomed to analyzing data using AI models. Typedef’s advanced AI data infrastructure manages the complexities associated with mixed AI workloads, such as context windows and token limits served through a clean user interface with the APIs and tooling engineers already recognize.
Most importantly, Typerdef’s AI data infrastructure enables fast pipeline experimentation to determine the potential value of AI workloads before they reach the production phase. “Data complexities and flawed data inputs are common obstacles on the journey to AI-readiness,” said Typedef co-founder Kostas Pardalis. “AI and data teams want the same rigor and reliability they expect from traditional data pipelines.”
Typerdef is making it possible for data teams to streamline their online analytic processing workloads and execute complex agentic workloads with much better scalability and predictability. As a result, data teams can now deliver on the promises of AI to stakeholders.
Conclusion
Typerdef is 100% focused on accelerating the innovation and deployment of its advanced AI data infrastructure by making it publicly available. That is why it released a significant percentage of its technology on GitHub under the name Fenic with an open-source license, allowing developers to use it freely.
