Dublin startup UniVec launches a new type of AI model aimed at solving one of the industry’s key infrastructure problems

Dublin startup UniVec introduces embedding interoperability models to reduce AI migration costs and infrastructure lock-in.

DUBLIN, DUBLIN, IRELAND, May 21, 2026 /EINPresswire.com/ -- UniVec, a Dublin-based AI infrastructure startup, today announced the release of its first embedding translation models:  a new class of AI models designed to make incompatible AI systems interoperable.

As companies race to build AI-powered search, recommendation engines, copilots, and agents, embeddings have quietly become one of the core infrastructure layers behind modern AI. But there’s a problem: every embedding model creates its own semantic space, effectively locking companies into the provider and model they started with.

Changing models often means re-embedding entire datasets, rebuilding retrieval systems, and accepting downtime, degraded search quality, or significant infrastructure cost.

UniVec is attempting to change that.

Its models are designed to translate vectors generated by one embedding model into another compatible semantic space, allowing companies to switch models, upgrade providers, consolidate systems, or experiment with newer AI infrastructure without rebuilding everything underneath.
The company describes the problem as “AI infrastructure lock-in” and believes embedding interoperability will become a foundational layer of the next generation of AI systems.

UniVec was founded in Dublin by Wade and Adrian Mihai, who previously worked together building large-scale AI systems at Opening.io and later at iCIMS.
Alongside the launch, the company also open-sourced several of its embedding translation models, positioning the move as part of a broader push toward open interoperability standards in AI infrastructure.

The launch comes as enterprises increasingly struggle with rapid model deprecations, rising vector storage costs, and pressure to support multiple AI providers simultaneously.

The company says it is actively expanding support for additional embedding models and exploring future interoperability challenges emerging across the rapidly evolving AI stack.

More information is available at univec.ai.

Andreea Wade
UniVec
https://univec.ai/contact
Visit us on social media:
LinkedIn

Legal Disclaimer:

EIN Presswire provides this news content "as is" without warranty of any kind. We do not accept any responsibility or liability for the accuracy, content, images, videos, licenses, completeness, legality, or reliability of the information contained in this article. If you have any complaints or copyright issues related to this article, kindly contact the author above.