Deals · Enterprise AI

Kyndred raises €500k Pre-Seed to build AI companions that remember you

Builds AI companion characters with a face, voice and persistent memory that interact with users in real time and retain context across conversations — relationship-based AI rather than transaction-based AI assistants.

ProYarn Desk · September 29, 2026Read this in French

in 𝕏 ✆

“Most AI characters today are really just text in a chat window. We build characters with a face, a voice and a memory, who respond to you in real time and remember what you told them. Company, not conversation. Our ambition is for your character to help you become a better version of yourself: to encourage you to get out of the house, see people and get organized.”

Alex Bararu — Co-founder and CMO, Kyndred

Kyndred (Romania) has raised €500,000 in a Pre-Seed round from Early Game Ventures. The company builds AI companions — characters with a face, voice, and persistent memory that interact with users in real time and retain context from previous conversations. Its first product, Maya, launched three months ago and has 150 paying customers.

What distinguishes a companion from an assistant

Co-founders Alex Bararu (CMO) and Rares Mosescu are building a deliberate architectural contrast to the AI assistant model. An assistant is optimised for task completion: the user opens it, extracts information or completes a task, and closes it. Kyndred's argument is that a companion is optimised for a different dynamic — the user returns not to complete a specific task but because there is an accumulated relationship.

Maya's design follows from this: a visual character rather than a text interface, voice interaction in real time, and memory that carries forward from session to session. The system learns a user's habits and preferences over time and is designed to assist with everyday activities — not in the sense of answering questions, but in the sense of providing consistent encouragement and organisation support.

Dan Calugareanu, Partner at Early Game Ventures, articulated the investor thesis: "Assistants are built on transactions: you open one, extract the information you need, and close it. Companions are built on relationships: you come back to them, and you do so without necessarily having a specific task to complete."

First traction

Kyndred generated its first revenue through a pre-launch campaign, before Maya was publicly available. The founders already had a community from previous entrepreneurial activity, and animations of the product attracted paying interest ahead of launch. Within three months of commercial availability, Maya has 150 paying customers.

Pre-Seed capital will fund improvements to Maya, development of additional AI companion characters, customer base growth, and team expansion — the company is currently hiring an AI Engineer and a Product Engineer.

Why it matters to investors

The AI companion category is crowded with consumer apps that have had sharp initial growth followed by high churn — users try the product, find it novel, and then stop returning. Kyndred's thesis is that the relationship model fixes the retention problem: a companion that accumulates memory of its user's preferences, habits, and history has an intrinsic switching cost that a transactional assistant does not. 150 paying customers three months after launch is a signal, not proof — the question is six-month and twelve-month retention, and whether the characters Kyndred plans to add after Maya create a portfolio dynamic or dilute the intimacy of any one relationship. Early Game Ventures is the right backer for a Romanian pre-seed: the firm has a track record of backing Eastern European consumer tech at the earliest stages and the network to support both team growth and a potential Central/Eastern European go-to-market path before a US push.

Sources

Pre-Seed Enterprise AI Romania Early Game Ventures

Threaded to this story

Enterprise AI · September 23, 2026

Every European round, in your inbox by 8am.

The day's seed and Series A rounds across France and Europe — threaded, sourced, and read in two minutes. Free.